EASTERBROOK, Circuit Judge.

Among the products of Baxter Healthcare Corp. are blood plasma derivatives, useful to promote blood clotting and provide antibodies against disease. Federal regulations require that incoming units of plasma be tested for hepatitis and human immunodeficiency virus before it is commingled with a larger pool from which commercial blood plasma products are prepared. 21 C.F.R. §§ 610.40, 610.45, 640.67. Joan P. Luckey, a former employee of Baxter Screening Laboratory (the division of Baxter Healthcare that tests plasma), filed this qui tam action in the name of the United States under the False Claims Act, 31 U.S.C. §§ 3729(a)(1), 3730(b), contending that Baxter falsely represented that it had tested incoming plasma. Luckey also sought relief in her own right, claiming that she had been dismissed in violation of § 3730(h) for exposing Baxter’s false claims. The district court granted summary judgment in Baxter’s favor. 2 F.Supp.2d 1034 (N.D.I11. 1998).

Blood plasma is clear. This tempts outside vendors to cheat — to palm off saline solution as plasma, or to dilute plasma with saline to sell a larger volume. On some occasions saline contamination is accidental, for saline solution is used to clean the machinery used to extract plasma from whole blood. Normal tests for hepatitis virus and hiv give meaningless results when applied to pure saline solution, and the presence of saline contamination in blood plasma leads to false negatives in some cases. Luckey believes that Baxter should use a total protein test to determine whether each incoming plasma unit contains saline solution. Managers at Baxter Screening Laboratory disagreed with Luckey’s proposal. Testing is costly, and the managers thought that saline dilution is not a sufficiently common problem that a separate test always is justified. Blood plasma is pooled before being packaged for sale, and a total protein test is administered to the pool. Thus there is no risk that saline will cause any shipping product to defeat the tests for hepatitis virus or hiv (tests that are performed on the pool at the end of processing, just as they are performed on each unit of plasma at the outset). Baxter also applies viral inactivation processes to each pool, reducing risks to a negligible level for the final product.

None of the federal regulations requires a supplier to administer a total protein test to each incoming unit. Nonetheless, the regulations do demand testing for hepatitis and hiv; to administer ineffective tests is the same as to administer no tests, Luckey contends. A representation that blood plasma has been tested, when no (proper) tests have been performed, is fraud on the federal government, Luckey submits, and therefore is actionable in a qui tam suit. Baxter denies that it makes any representation to the United States about its testing; when it sells plasma products to federal agencies, its only relevant statements *732 are that it holds a license from the Food and Drug Administration and that the products are free of hepatitis and hiv. But we shall assume for current purposes that Luckey could show that a representation about testing is made indirectly. The assumption does not assist her.

The principal difficulty with Luckey’s position is its equation of ineffective testing with no testing, followed by the proposition that a battle of experts in litigation determines whether testing is “effective.” Only in Humpty Dumpty’s world of word games would anyone apply the label “fraud” to the kind of representations Baxter made. Perhaps Luckey could persuade the Food and Drug Administration to require more or different testing; but when a supplier complies with the existing regulations, it is entitled to represent to the government (and the world) that it has done so, without facing a claim of deception.

What tests to perform, and when, are difficult questions. No one doubts that, when saline contamination is a possibility, a total protein test reduces the numbers of false negatives from other tests. Similarly, however, no one believes (at least, no one should believe) that every possible test, no matter how expensive, should be administered to avoid every conceivable risk, no matter how small. Some form of cost-benefit inquiry must be carried out. See Stephen Breyer, Breaking the Vicious Circle: Toward Effective Risk Regulation (1993). Are the extra costs worth the extra expense? That depends not only on the volume of saline-contaminated plasma received, and the probability that saline will lead to false negative results from standard tests, but also on the effectiveness of the tests that are applied to the plasma after pooling. Tests performed on the pool have few if any false negatives. Thus the real cost of not using a total protein test earlier is that some plasma with hepatitis or hiv will get into a larger pool and cause the rejection of the whole pool, wasting quantities of plasma that were sound when received. Waste is regrettable, but the cost of thrown-out plasma may be measured in dollars; and if Baxter is willing to bear the expense (thinking the cost of rejected pools to be less than the cost of more comprehensive testing earlier), it is hard to see why the federal government should force it to use the total protein test on each unit — and absolutely impossible to see why Baxter’s resolution of the testing issue should be called “false” or “fraudulent.” Most companies in Baxter’s position will have scientists who argue for additional precautions, and others who think that the existing precautions are enough (if not excessive). Luckey, a lab technician, was in the more-precautions camp. As written, the regulations do not resolve this debate. Baxter had to make a decision for itself. Since 1995 Baxter has subjected some incoming units of plasma to total protein tests, but this change in its testing protocol does not imply that representations about the former protocol were false.

Equating “imperfect tests” with “no tests” would strain language past the breaking point. True enough, a claim can be false or fraudulent if the speaker offers a misleading half-truth. Thus the assertion “we performed all appropriate tests” could be called false if Baxter meant by “appropriate” only “required by regulation,” while knowing that the test results were worthless and the product defective. What keeps this approach from swallowing the norm that literal truth is enough is the requirement that the plaintiff show not only that the omitted facts were material to the listener’s decision, see Neder v. United States, — U.S. -, -, 119 S.Ct. 1827, 144 L.Ed.2d 35 (1999), but also that the speaker intend the statement or omission to deceive, Ernst & Ernst v. Hochfelder, 425 U.S. 185, 96 S.Ct. 1375, 47 L.Ed.2d 668 (1976). See 31 U.S.C. § 3729(b). Luckey has not offered any evidence tending to show that the omission *733 of a total protein test at the plasma intake stage (before pooling) was material to the United States’ buying decision; the Department of Justice has conspicuously declined to adopt Luckey’s position or to prosecute this claim on its own behalf. As far as this record reveals, the federal government is 100% satisfied with the blood products it receives from Baxter and with the representations made in connection with the sales. Moreover, there is no evidence that Baxter intended to deceive anyone about what it was doing, or the validity of the test results. All this record reveals is a dispute about whether Baxter’s testing protocols could be improved. An affirmative answer to that question would not suggest that Baxter’s representations to the United States in years past were false or fraudulent. United States ex rel. Lamers v. Green Bay, 168 F.3d 1013, 1019 (7th Cir.1999), holds that technical violations of a federal regulation on which a claim is based do not make the claim “false”; in this case we do not even have an established violation.

As for the claim under § 3730(h): this statute prohibits an employer from discharging or taking other adverse action against an employee “because of lawful acts done by the employee on behalf of the employee or others in furtherance of an action under this section, including investigation for, initiation of, testimony for, or assistance in an action filed or to be filed under this section”. The district court granted summary judgment to Baxter because Luckey could not show any causal link between this suit and her discharge. Baxter fired Luckey in April 1995, before her qui tam suit was unsealed and thus before Baxter learned of the complaint. We held in Neal v. Honeywell Inc., 33 F.3d 860 (7th Cir.1994), that § 3730(h) prevents retaliation against activities preparatory to suit, but it was established in the district court that Baxterwas not aware of any such activities on Luckey’s part. She contends nonetheless that because Baxter knew her position that every unit of incoming blood plasma should be subjected to a total protein test, Baxter was aware of an “investigation for” suit, or at least the possibility of “assistance in” a suit to be filed. Not so. Intra-corporate debates about optimal testing protocols cannot be equated to knowledge of litigation — not unless taking one side of a medical or scientific dispute is “fraud,” a position we have already rejected. An employer is entitled to treat a suggestion for improvement as what it purports to be rather than as a precursor to litigation. Luckey’s discharge therefore cannot be understood as one “because of lawful acts done by the employee ... in furtherance of an action under this section”.

Even if Luckey’s statement to a coworker in March 1995 that she planned to “shut down the bsl” and “get rid of’ some supervisors can be taken as a hint that false-claim litigation (as opposed to sabotage) was under consideration, Baxter was not frozen in place. Neal observes that the employee who “fabricates a tale of fraud to extract concessions from the employer, or who just imagines fraud but lacks proof, legitimately may be sacked.... [Ejmployees who use reports of fraud to better their own position, or who behave like Chicken Little, impose costs on employers without advancing any of the goals of the False Claims Act.” 33 F.3d at 864. Saber-rattling is not protected conduct. Only investigation, testimony, and litigation are protected, and none of these led to Luckey’s discharge.

Finally, Luckey contends that, because she is now employed at a much lower wage than formerly, the district judge should not have required her to pay the costs of litigation, which came to about $20,000. Appellate review is deferential, and it is impossible to say that the district judge abused his discretion, given that “costs other than attorneys’ fees shall be allowed as of course to the prevailing party”. Fed.R.Civ.P. 54(d)(1). We have held that an award of costs is proper even when the unsuccessful litigant had been allowed *734 to proceed in forma paupens, as Luckey was not. McGill v. Faulkner, 18 F.3d 456 (7th Cir.1994). Someone has to bear the costs of litigation, and the winner has much the better claim to be spared them— not just a morally or economically better claim, but under Rule 54(d) a legally better claim. A district judge might exercise his discretion to require each side to bear its own costs when the case is close, or when the side with the better position loses on a technicality. This case, however, was not close; it verges on being frivolous. Straitened circumstances do not justify filing weak suits and then demanding that someone else pay the bill.

Affirmed.