Opinion for the Court filed by Senior Circuit Judge WILLIAMS.

WILLIAMS, Senior Circuit Judge:

Before the terrorist attacks of September 11, 2001, commercial airlines exercised responsibility for screening passengers and property at U.S. airports. Shortly after the attacks, Congress passed the Aviation and Transportation Security Act (“ATSA”), Pub.L. No. 107-71, 115 Stat. 625 (2001), establishing the Transportation Security Administration (“TSA”) and entrusting it with the primary responsibility for civil aviation security. 49 U.S.C. § 114. We deal here with several airlines’ arguments that TSA has erroneously over *1068 charged them for their statutory portion of these security costs.

The ATSA authorizes TSA to impose two types of fees to fund its security services. The first, which is not at issue here, is a fee on airline passengers. 49 U.S.C. § .44940(a)(1). The second type of fee— referred to as the Aviation and Security Infrastructure Fee (“ASIF”), 49 C.F.R. § 1511.1(b) — is imposed directly on airlines. It is meant to plug the gap between the costs of TSA’s civil aviation security services and the sums raised by the passenger fee, but it is subject to two important limits. 49 U.S.C. § 44940(a)(2)(A). Petitioners — -a group of 22 airlines. — are claiming that TSA improperly subjected them to approximately $98 million a year in increased ASIF liabilities.

ATSA’s two limits on fees are its “overall” and its “per-carrier” limits. Under the overall limit, the fees in each fiscal year “may not exceed, in the aggregate, the amounts paid in calendar year 2000 by carriers ... for screening passengers and property, as determined by the Under Secretary.” 49 U.S.C. § 44940(a)(2)(B)(i). Under the per-carrier limit, the fees collected from a carrier for fiscal years 2002, 2003 and 2004 “may not exceed the amount paid in calendar year 2000 by that carrier for screening passengers and property, as determined by the Under Secretary.” 49 U.S.C. § 44940(a)(2)(B)(ii) (emphasis added). Starting with fiscal year 2005, the act allows the Under Secretary to determine the per-carrier limit “on the basis of market share or any other appropriate measure in lieu of actual screening costs in calendar year 2000.” 49 U.S.C. § 44940(a)(2)(B)(iii).

In its implementing regulations, TSA required every covered carrier to submit a form — referred to as “Appendix A” — detailing its passenger and screening costs for the year 2000. 49 C.F.R. § 1511.5(d). It also required carriers to provide an audit of their reported costs. Id. § 1511.9. For the years 2002-2004, TSA generally set each carrier’s annual fee at the level of costs listed in the carrier’s Appendix A.

In 2004, with the Department of Homeland Security Appropriations Act, Pub.L. No. 108-334, 118 Stat. 1298 (2004) (the “2004 DHS Appropriations Act”), Congress intervened. to make sure that TSA was collecting its full entitlement under ATSA. It directed the Government Accountability Office (“GAO”) to review the airlines’ cost information for the year 2000. And it stated that, beginning with amounts due in the year 2005, if “the result of this review is that an air carrier or foreign air carrier has not paid the appropriate fee to the Transportation Security Administration ..., the Secretary of Homeland Security shall undertake all necessary actions to ensure that such amounts are collected.” Id.

The GAO’s report concluded that the airlines had collectively under-reported their security screening costs in the year 2000 by an estimated $129 million. United States Government Accountability Office, Aviation Fees: Review of Air Carriers’ Year 2000 Passenger and Property Screening Costs 7 (2005).

Relying in part on the GAO’s analysis, TSA determined that each of petitioners had under-reported its year 2000 screening costs. It began by calculating the industry’s average cost per passenger screened. It then compared that average with each airline’s reported cost per passenger screened. For those airlines whose reported costs were at or above the industry average, it assessed no additional liability. It also gave a pass to airlines whose reported costs were, below the industry average, but that had presented an adequate audit of their reporting costs. *1069 When TSA determined — as it did for all of the petitioners — that an airline reported below-average costs and did not provide an adequate audit, TSA presumed that its screening costs per passenger in the year 2000 were equal to the industry average. It then calculated the airline’s total screening costs for 2000 by multiplying the industry average by the number of passengers that airline screened in 2000. Finally it estimated each airline’s additional ASIF liability by subtracting its reported year 2000 costs from the new figure.

TSA’s director of revenue sent each petitioner a letter describing this method and advising the petitioner of its additional liability. Petitioners appealed to TSA, which upheld the initial decisions in all relevant respects.

In their joint brief, petitioners challenge TSA’s final decisions, claiming that they reflected substantive statutory violations, were arbitrary and capricious, and were flawed procedurally. We find merit in the attack on TSA’s understanding of the ATSA’s “overall” limit, but not in the other objections.

In addition, three of the petitioners advance individual claims, one of which (that of Spirit Airlines (“Spirit”)) prevails.

Before reaching the merits, we need to address the effect of two ATSA provisions for jurisdiction-stripping. The original ATSA provided that “[determinations of the Under Secretary under this subpara-graph [i.e., 49 U.S.C. § 44940(a)(2)(B), stating the limitations on air carrier fees] are not subject to judicial review.” Pub.L. No. 107-71, 115 Stat. 597, 625 (2001). In the Consolidated Appropriations Act, P.L. 110-161, § 540, 121 Stat. 1944 (December 26, 2007), Congress relaxed this restriction, creating an exception for “estimates and additional collections made pursuant to the appropriation for Aviation Security in Public Law 108-334 [i.e., collections made pursuant to the 2004 DHS Appropriations Act]: ... Provided ... That such judicial review shall be limited only to additional amounts collected by the Secretary before October 1, 2007.” 49 U.S.C. § 44940(a)(2)(B)(iv). As the collections here within the scope of the jurisdiction-stripping provision were made pursuant to the directive of the 2004 DHS Appropriations Act, and all the issues apply in part to amounts collected by TSA before October 1, 2007, we have jurisdiction over all issues that petitioners pose. Because some of the issues fall outside the basic jurisdiction-stripping clause, we will note below- — as to each issue on which petitioners prevail — whether our judgment applies to collections on or after October 1, 2007.

“Overall limit.” The ATSA’s “overall limit” provides that the fees in each fiscal year “may not exceed, in the aggregate, the amounts paid in calendar year 2000 by carriers ... for screening passengers and property.” 49 U.S.C. § 44940(a)(2)(B)® (emphasis added). Petitioners argue that TSA violated the plain language of the provision by basing its calculation of the fees on a GAO estimate which had included the costs of screening wow-passengers, such as “meeters-andgreeters” and sightseers. TSA acknowledges inclusion of the costs of screening such individuals, but seeks to justify doing so. Its arguments do not convince us.

TSA argues that the reference to “screening passengers” is ambiguous, that the word “screening” may mean something more than the simple evaluation of whether a passenger poses a threat to aviation security. “To screen” may also mean “to protect.” (“The mother screened her child from the pounding hailstones.”) There *1070 fore, TSA suggests, the phrase “screening passengers” can be read to include anything done to protect passengers, which of course would include exerting control over the access of potentially dangerous non-passengers. But “[ajmbiguity is a creature not of definitional possibilities but of statutory context.” Brown v. Gardner, 513 U.S. 115, 118, 115 S.Ct. 552, 130 L.Ed.2d 462 (1994). In the context of airport security, the phrase “screening passengers” has a widely understood meaning: it refers to the process of searching airline passengers at an airport security checkpoint, not to the entire set of activities undertaken to promote passenger safety.

TSA also notes that the statute refers not merely to screening passengers, but also to screening “property.” True enough; and the statute does not limit the relevant “property” to that of passengers. Thus, TSA could include the costs of screening the property of non-passengers in its calculation. But that authority provides it no justification for also including the costs of screening the non-passengers themselves.

TSA also asserts an equally unpersuasive argument from statutory purpose. It asserts that Congress did not intend to “bestow a windfall on carriers by assuming the full burden of providing services that inured to the benefit of the industry, and that had previously been provided at the industry’s expense.” Respondent’s Br. at 39. The argument might conceivably trump the statutory language if it accorded with the facts, but it doesn’t. As petitioners observe, sightseers and meeters-and-greeters “have been barred from secure airport areas since the September 11 attacks and, therefore, are not screened by TSA.” Petitioners’ Reply Br. at 11. TSA’s interpretation serves no windfall-prevention purpose.

Finally, TSA argues that petitioners’ interpretation is inconsistent with another part of the statute. It points to 49 U.S.C. § 44901(a), which directs TSA to “provide for the screening of all passengers and property.” TSA suggests that this provision must be interpreted to give it the power to screen all persons who enter controlled boarding areas; it would be absurd to deny TSA power to screen non-passengers. And if “passengers and property” includes non-passengers in § 44901(a), TSA argues, the phrase should have the same meaning in § 44940(a)(2)(B)®. Supporting TSA’s theory is the standard presumption that “identical words used in different parts of the same act are intended to have the same meaning.” Sorenson v. Sec’y of Treasury, 475 U.S. 851, 860, 106 S.Ct. 1600, 89 L.Ed.2d 855 (1986) (quoting Hel-vering v. Stockholms Enskilda Bank, 293 U.S. 84, 87, 55 S.Ct. 50, 79 L.Ed. 211 (1934)). But this presumption is not so strong as to displace the plain meaning of a statutory provision simply by virtue of the fact that interpreting a different provision the same way would or might be absurd. In addition, although the question of § 44901(a)’s meaning is not before us, we note that we are not convinced that it is the only possible source of TSA’s power to screen non-passengers. See 49 U.S.C. § 44903(h) (authorizing TSA screening of “all individuals” before entry into a secured area of covered airports); 49 C.F.R. §§ 1540.105, 1540.107 (including 49 U.S.C. § 44903 among sources of authority for certain airport screening activities). In short, TSA violated the plain meaning of the ATSA’s overall limit when it included the costs of screening non-passengers in its estimate of the costs of covered screening in 2000.

This holding governs amounts collected before October 1, 2007; what of the effect *1071 of the jurisdiction-stripping provision on amounts collected thereafter? Our conclusion rests not on a review of a “determination ... under” the subparagraph covered by the provision, but rather resolves the question whether TSA has made the kind of determination required by the statute. We drew just this distinction in COMSAT Corp. v. FCC, 114 F.3d 223 (D.C.Cir.1997), which involved a provision precluding judicial review of “increases or decreases in fees made by amendments pursuant to this paragraph.” Id. at 224 (quoting 47 U.S.C. § 159(b)(3)). We read the clause to mean simply that “the courts may not review the Commission’s actions where the Commission has acted within the scope of its authority” under the controlling statute. Id. at 227. So, too, here. Therefore, the jurisdiction-stripping provision does not apply, and our holding on this point governs the collections made after 2007.

“Per-carrier limit.” Petitioners’ argument here is far less persuasive. They assert that the statutory language— which limits each carrier’s ASIF to “the amount [it] paid in calendar year 2000 ... for screening passengers and property, as determined by the Under Secretary,” § 44940(a)(2)(B)(ii) — unconditionally precludes TSA’s use of the industry average as a proxy for petitioners’ cost per passenger screened. We see no such preclusion.

First, the statute’s “amount paid” language is qualified: “as determined by the Under Secretary.” As then-judge Roberts observed in AFL-CIO v. Chao, 409 F.3d 377 (D.C.Cir.2005), ‘We have noted in the past the ‘distinction between the objective existence of certain conditions and the Secretary’s determination that such conditions are present,’ stressing that a statute phrased in the latter terms ‘fairly exudes deference’ to the Secretary.” Id. at 393 (Roberts, J., concurring in part and dissenting in part) (quoting Kreis v. Sec’y of Air Force, 866 F.2d 1508, 1513 (D.C.Cir.1989)). The ATSA similarly gives TSA broad discretion to choose a suitable method for making the required determination.

Here TSA’s choices were clearly permissible. Although the airlines may prefer that TSA rely on their Appendix A information, TSA was certainly entitled to conclude that, in the absence of an audit, such data were not reliable enough. TSA was also free to select a reasonable alternative, such as the industry’s average cost. That average cost — multiplied by a logically chosen carrier-specific variable, the number of passengers screened by the carrier in the year 2000 — in fact constitutes a measurement of a specific carrier’s screening costs.

The 2004 DHS Appropriations Act called for GAO review of “the calendar year 2000 cost information for screening passengers and property pursuant to section If.h9h0(a)(2),” 118 Stat. at 1303 (emphasis added); petitioners therefore reframe their per-carrier argument to claim that the 2004 Act requires the GAO to examine the individual air carriers’ “actual” security costs, instead of using sampling data. Petitioners’ Br. at 38, 40. The argument fails here, naturally, for the same reasons as it did in the context of § 44940(a)(2)(B) itself.

Starting with the fiscal year 2005, of course, 49 U.S.C. § 44930(a)(2)(B)(iii) gives TSA even broader latitude, authorizing application of the per-carrier limit by reference to “market share or any other appropriate measure.” For the collections at issue here, however, TSA hasn’t invoked this section.

Finally, we note that the per-carrier limit, like the overall limit, rests on an estimate of the costs of “screening passengers and property.” Because TSA’s industry' average included the costs of screening *1072 non-passengers, that calculation was not a “determination ... under” 49 U.S.C. § 44940(a)(2)(B), and must be corrected on remand.

Claims that TSA acted arbitrarily and capriciously. Petitioners’ first such argument is that TSA penalized them for not complying with its requirement of providing an unqualified audit opinion for their Appendix As — a requirement they say was impossible to fulfill. Their records were inadequate to provide the required information without making significant assumptions, and thus their auditors would not provide unqualified opinions.

Assuming arguendo that TSA was obligated to supply a feasible alternative before relying on an industry average, this argument still fails because TSA did in fact supply such an alternative. In response to the industry’s concerns, TSA announced that it would accept a qualified opinion if an auditor could not provide an unqualified one. But TSA stipulated that, as the qualified opinions would not supply adequate “details and reasoning,” a carrier relying on such an opinion would have to submit the auditor’s working papers (rather than merely assuring their availability to TSA). See Joint Appendix (“J.A.”) 365-67.

Petitioners do not claim that they complied with this alternative procedure. In their reply brief, however, they seem to argue that they could not have done so, on the grounds that an auditor’s working papers are its property, not theirs. Petitioners’ Reply Br. at 15. It is not clear to us why an auditor’s property interest (presumably a negotiable matter in any event) would allow the airlines to make the papers available upon request, but not to submit them in the way outlined by TSA. In any event, petitioners waived the argument by failing to make it in their opening brief. Carducci v. Regan, 714 F.2d 171, 177 (D.C.Cir.1983).

In petitioners’ second arbitrary-and-capricious argument, they reformulate their contention that TSA misread the statute’s per-carrier limit, now saying that it was arbitrary to use the industry average cost per passenger as a proxy for each petitioner’s per-passenger cost. They also say that TSA wrongly ignored their Appendix As and failed to give proper weight to their evidence that their screening costs were below the industry average.

Assuming this argument is actually distinct from the statutory interpretation claim, it fares no better. Given the technical character of the issue, TSA is entitled to a good deal of deference on its resolution. See, e.g., West Virginia v. EPA, 362 F.3d 861, 867 (D.C.Cir.2004); Milk Train, Inc. v. Venenan, 310 F.3d 747, 754 (D.C.Cir.2002). Here, the carriers’ own explanations as to why it was impossible for them to obtain unqualified audits completely undermine their position: for carriers that did not collect adequate carrier-specific information on screening costs for 2000, it was hardly arbitrary for TSA to rely on an industry average.

Petitioners then object that the GAO report on which TSA relied had allocated excessive amounts of airport law enforcement officer (“LEO”) costs to passenger screening. (The calculations, of course, were based on TSA’s mistaken decision to include the costs of non-passenger screening; it may be that the redo necessitated by our decision will occasion agreement between the parties on these subsidiary issues, or at least less disagreement.) Specifically, when an airport explicitly identified an LEO charge as being for “Flexible Response,” GAO, in reliance on its contractor, included 100% of LEO costs in screening; when an airport allocated a portion of its airport-wide LEO budget to “the terminal cost center,” GAO accepted *1073 its contractor’s decision to “judgmentally” apply 50% of such costs to screening. See J.A. 15-16(GAO), 108 (contractor).

These allocations are hardly rock solid. But TSA was operating in a data-poor environment, as shown by petitioners’ own arguments about their carrier-specific data. TSA’s contractor interviewed airport officials, eliciting a wide range of statements about the actual role of LEOs. This is most certainly not a case where there is reason to believe that materially superior information could be readily obtained, and where, accordingly, a court would be likely to find an agency’s disregard of the alternative as arbitrary. Air Transport Ass’n of Canada v. FAA, 323 F.3d 1093, 1095-97 (D.C.Cir.2003). Any decision here would have required considerable guesswork, and we cannot say that TSA’s guesses were unreasonable.

Petitioners also claim that the GAO report relied on faulty statistical techniques. They argue, for instance, that in estimating the cost of private screening contractors, the GAO relied on data from only nine companies, which were not shown to be representative of the field as a whole. In fact GAO’s contractor has sought data from the 10 largest screening contractors, which represented between 84% and 95% of the market (depending on whether one relies on number of screeners or number of passengers screened). J.A. 72. It excluded data from the largest simply because nearly 7000 of its invoices appeared to be incomplete and inconsistent. J.A. 79. While complaining about GAO’s reliance on “such small samples,” Petitioners’ Br. at 58, petitioners quite sensibly make no argument for inclusion of data from the largest screening company. Nor do they acknowledge the contractor’s finding that the 2d-through-10th largest firms accounted for between 63% and 67% of all screening. J.A. 80.

TSA accepted the results, observing that, while other approaches to data collection and statistical analysis were available, they would not have yielded results that were “substantially different or more reliable.” J.A. 565, 566. Even before us, petitioners do not directly claim the existence of an alternative so visibly superior that TSA could be faulted for making the choice it did. We cannot say the choice was arbitrary or capricious.

Petitioners similarly object to the GAO’s extrapolation of other airport costs for all 419 airports from a sample of 59. (In fact GAO sought data from 70 airports, the top 20 and 50 others. O’Hare did not provide sufficient information for inclusion, but the other top 19 did.) The 70 sampled airports accounted for about 75% of screened passengers in 2000. J.A. 15, 98.

They argue that the extrapolation was unreliable because it was based on screened passenger volumes, which do not correlate closely with actual airport costs. TSA found, however, that this objection was invalid because the formulation used by the GAO was not directly tied to the volume of the passengers screened. J.A. 567. More generally, the TSA found that the lack of independently verifiable data forced the GAO to make assumptions based on its “professional judgment and expertise” as well as the available data. J.A. 566. Although these assumptions can be questioned, it was not arbitrary and capricious for the TSA to rely on them.

Finally, petitioners argue that the decisions were arbitrary and capricious because TSA violated its own regulations, specifically 49 C.F.R. §§ 1511.11(a), 1511.5(c) & (d), which petitioners try to read as mandating individual company audits by TSA rather than reliance on GAO. A glance at the regulations shows that they impose no such mandate.

*1074 Procedural claims. Petitioners claim a right under the due process clause to see certain documents relevant to TSA’s decisions, such as the data file used by the GAO. TSA responded to this claim with an assertion that the airlines’ access to the GAO report and their own records should be “sufficient to substantiate all relevant grounds, if any, for relief’ from TSA’s claims for additional compensation. J.A. 373. (The airlines also, of course, had the GAO contractor’s report, with considerable additional detail.)

The airlines correctly cite McClelland v. Andrus, 606 F.2d 1278, 1285 (D.C.Cir.1979), for the general proposition that in some circumstances due process will entitle a party to discovery in an agency proceeding. There we appeared to assume that the decision was adjudicative, and thus belonged to the class of cases for which due process is typically, and almost exclusively, applicable. See Decatur Liquors, Inc. v. District of Columbia, 478 F.3d 360, 363 (D.C.Cir.2007); Wisconsin Gas Co. v. FERC, 770 F.2d 1144, 1166 & n. 35 (D.C.Cir.1985). Compare Londoner v. Denver, 210 U.S. 373, 385, 28 S.Ct. 708, 52 L.Ed. 1103 (1908), with Bi-Metallic Inv. Co. v. State Bd. of Equalization, 239 U.S. 441, 445-46, 36 S.Ct. 141, 60 L.Ed. 372 (1915). Moreover, as a later decision recognized, “McClelland was seeking a specific document ‘uniquely relevant to [his] case.’ ” EchoStar Communications Corp. v. FCC, 292 F.3d 749, 756 (D.C.Cir.2002). The document requests here appear both less specific and less urgent than McClelland’s. In EchoStar, in fact, we found no due process violation even though the agency’s explanation was “terse” and “cryptic.” Id. at 755, 756. Given the nature of TSA’s decision — an inquiry into industry-wide costs (once TSA validly decided to rely on such data rather than on petitioners’ Appendix As) — and the “extreme deference” with which we review agency denials of discovery, id. at 756, the denial here clearly passes (again assuming the application of due process requirements at all).

Petitioners also object to contacts between the official who rendered the final decisions and those responsible for the initial decisions, contacts that we may assume would violate the separation-of-functions rules of the Administrative Procedure Act (“APA”) if those rules were applicable. See 5 U.S.C. § 554(d). But § 554 applies only to “adjudication^] required by statute to be determined on the record after opportunity for an agency hearing,” 5 U.S.C. § 553(a), and nothing in the ATSA or the DHS Appropriations Act requires such a hearing. Cf. Dist. No. 1., Pac. Coast Dist., Marine Eng’rs Beneficial Ass’n v. Maritime Admin., 215 F.3d 37, 42 (D.C.Cir.2000) (addressing claim of ex parte contacts in matter not covered by the APA’s ban).

Petitioners also assert a constitutional theory, resting primarily on Stone v. FDIC, 179 F.3d 1368 (Fed.Cir.1999), a case involving discharge of a government employee found to have a property interest in his job and thus clearly entitled to due process. Id. at 1374-75. But the court in Stone expressly confined its opinion to instances where the decider received “new and material” information, of which there is no claim, nor any reason to suspect, here. Thus the Stone decision appears quite consistent with Withrow v. Larkin, 421 U.S. 35, 58, 95 S.Ct. 1456, 43 L.Ed.2d 712 (1975), rejecting any due process requirement of separation of functions unless special circumstances indicate “that the risk of unfairness is intolerably high.” Even if we assumed in petitioners’ favor that the proceeding was subject to due process requirements at all, we would not have the rare conditions rendering the *1075 agency’s procedures unconstitutional. See also Gottlieb v. Pena, 41 F.3d 730, 737 (D.C.Cir.1994) (upholding against due process challenge an adjudicative procedure allowing staff to communicate ex parte with the ultimate decisionmaker); Chem. Waste Mgmt. Inc. v. EPA, 873 F.2d 1477, 1484 (D.C.Cir.1989) (relying on Withrow v. Larkin to reject facial attack on regulations allowing combination of functions).

Finally, petitioners argue that they had a due process right to an oral hearing. Again, assuming that the proceeding was of a nature to make due process requirements applicable, the claim fails because of the nature of the issues. Here they involved statutory construction, the validity of GAO’s statistical methods, and the accuracy of the carriers’ cost information— issues of a kind that can be adequately resolved on written submissions. See Lomak Petroleum, Inc. v. FERC, 206 F.3d 1193, 1199 (D.C.Cir.2000); CNG Transmission Corp. v. FERC, 40 F.3d 1289, 1295 (D.C.Cir.1994). Petitioners rely primarily on Gray Panthers v. Schweiker, 652 F.2d 146, 170 (D.C.Cir.1980), a healthcare benefits case. There we held: “Where factual issues involving the credibility or veracity of the claimant are at stake, particular consideration of a policy granting on request an oral interview before the final denial on reconsideration should be given.” Id. at 172 (emphasis added). The difference in the character of issues could hardly be more stark.

Individual airline claims. We close with the individual claims of Spirit, American Airlines (“American”) and Northwest Airlines (“Northwest”), in that order.

TSA acknowledged that the language of the audit submitted by Spirit complied with the requirements of the relevant regulation, 49 C.F.R. § 1511.9, yet declined to classify the opinion as “unqualified.” It explained that the audit did not include information that was “further required by clarification published in the Federal Register,” citing 67 Fed.Reg. 21,582, 21,584 (May 1, 2002). It therefore subjected Spirit to additional ASIF liability. J.A. 1824.

But the “clarification” to which TSA referred was contained in a guidance that explicitly states that it “does not impose any additional requirements” and that “[cjarriers should not infer that it represents the only acceptable means of completing Appendix A.” 67 Fed.Reg. 21,582, 21,582, 21583. Thus the guidance did not “further require” anything, and TSA’s stated grounds for rejecting Spirit’s audit do not hold water. We therefore set aside the additional ASIFs imposed on Spirit with respect to all amounts collected before October 1, 2007. As TSA’s error here is plainly a “determination ... under” 49 U.S.C. § 44940(a)(2)(B), however, the jurisdictional bar of 49 U.S.C. § 44940(a)(2)(B)(iv) applies to later collections.

American argues that TSA owes it almost $14 million for development and installation of an inline baggage security system. It asserts that under the common law right of offset, its ASIF should be reduced by the amount it is owed. The initial version of this opinion disposed of American’s claim by relying on 41 U.S.C. § 605, a statute not cited by the parties. American later petitioned for rehearing, arguing that our analysis of the statute was erroneous, but conceding that its claim was not validly presented in this proceeding. Ordinarily, we would not excise a legal analysis from an opinion simply on the *1076 basis of the losing party’s conceding the ultimate issue of another ground. In this case, however, we accept American’s concession rather than rely on a resolution that the parties never had an opportunity to satisfactorily address. Thus, assuming without deciding that a petitioner has the right to claim an offset from an agency before our court, American has not validly presented any such claim.

Finally, Northwest objects to being charged for screening costs that TSA did not assume. When calculating its total year 2000 costs for its Appendix A, Northwest excluded screening costs that it continued to bear. In 2005 TSA informed Northwest that it should have included those costs in its ASIF payment. As a result, quite independently of TSA’s substitution of its calculation of screening costs for the carriers’, Northwest was over $3 million in arrears and would have to pay higher fees going forward. J.A. 1548-49.

Northwest has a two-fold claim: it argues that statute prohibits TSA from imposing ASIFs based on security costs it has not taken over, and that in any case TSA was not authorized to impose these fees retroactively. Both arguments fail.

Northwest’s insuperable difficulty is that the limit that it seeks to impose — that an airline can only be charged for as much of its year 2000 screening costs as TSA has taken over- — -is nowhere to be found in the statute. We rehearsed at the outset the statute’s overall and per-carrier limits, and those are the only ones it states. It thus provides no protection for airlines against what may seem to be double collection for costs an airline continues to bear.

Nor is there anything retroactive — in any legally material sense — in TSA’s collection of the additional fees. TSA is not retroactively imposing a new rate; rather, it is collecting amounts that Northwest, because of its mistaken calculation, had failed to pay.

In sum, TSA erred in its interpretation of the ATSA’s overall limit and in its classification of Spirit’s audit opinion. We therefore remand for modifications consistent with this opinion, and otherwise affirm TSA’s decision in all respects.

So ordered.