POSNER, Chief Judge.

When this securities fraud case was last before us, we held that the district judge had erred when he granted the defendants’ motion to dismiss on the ground that it was apparent from the face of the complaint that the suit had been filed after the one-year statute of limitations applicable to suits under the SEC’s Rule 10b-5 had run. LaSalle v. Medco Research, Inc., 54 F.3d 443 (7th Cir.1995) (same case, different lead named plaintiff). We rejected the proposition “that the conjunction of optimistic forecasts -with a sharp drop in price establishes inquiry notice as a matter of law.” Id. at 447. On remand, the district judge again granted summary judgment for the defendants on the basis of the statute of limitations. We had left open the possibility that this might be appropriate if, when Medco’s stock price had been plummeting notwithstanding the optimistic forecasts, the prices of its competitors’ shares had been holding steady or rising. But in granting- summary judgment the second time the judge did not allude to these price movements. The defendants had on remand presented some evidence concerning them — but it was evidence that Medco’s stock price had moved in tandem with the prices of its competitors’ stocks. The defendants could not use this evidence to show that the statute of limitations had run, but they could and did use it to support an alternative ground for dismissal of the suit, to which the judge however did not allude, that there was no causal relation between the alleged fraud and the loss to the members of the plaintiff class. To support their statute of limitations defense, the defendants presented another kind of evidence, which the district judge found convincing — articles published in the trade press that should have made the plaintiffs suspicious.

It would have been helpful to us in this second round if the district judge had related his analysis of the new evidence to the principles that we had set forth in our opinion reversing his first decision to guide the proceedings on remand. The doctrine of law of the case requires the trial court to conform any further proceedings on remand to the principles set forth in the appellate opinion unless there is a compelling reason to depart, such as a controlling decision by the Supreme Court rendered after the appellate court’s decision. We are left to speculate whether the district judge thought he was following our lead or for some undisclosed reason had decided to strike out on his own. Appellate review of a decision to grant summary judgment is plenary, meaning that we don’t defer to the trial court’s decision. Nevertheless, it is a big help to the reviewing court to be told by the district judge how he thought his decision on remand followed from the appellate court’s previous ruling on the same issues in the same case.

When last the case was before us, the only basis upon which the defendants could argue that the statute of limitations had run was the narrative in the complaint. The suit had been filed on September 1, 1993, and the issue was therefore whether by September 1, 1992, the plaintiffs should have learned of the fraud. According to the complaint, in April of 1992 Medco had announced that its application to the Food and Drug Administration for approval of a new cardiac drug (Adenoscan) crucial to the company’s profitability was “on track.” A few weeks later the price of Medco’s stock, which had peaked at $31.75 in January, plunged to $15.25. In June, the FDA recalled several batches of a very simi *784 lar drug called Adenocard that was being manufactured in the same plant by the same firm, LyphoMed, that Medco had licensed to manufacture Adenoscan, the successor product to Adenocard. It was plausible to suppose that the problems that LyphoMed had encountered in manufacturing Adenocard might lap over to Adenoscan and delay the approval of that drug. In April of 1993, with Adenoscan still not approved, Medco announced that it was going to sue FujisawaUSA, LyphoMed’s parent. The suit was filed the next month, and revealed that the problems at the LyphoMed plant had caused Medco to withdraw temporarily its application for the approval of Adenoscan a week before it had told the investing public that the application was on track. On these facts — all we had to go on when the ease was last here — we held that not until April 1993 did investors have enough information to start the statute of limitations running; the suit filed in September was well within a year of that date.

The evidence presented by the defendants on remand consists primarily of articles published before September 1992 that posted “storm warnings,” in the district judge’s phrase (a cliché in opinions about investors’ diligence, e.g., Dodds v. Cigna Securities, Inc., 12 F.3d 346, 350 (2d Cir.1993)), of trouble ahead for Medco. One investment newsletter said in December 1990 that Med-co was “an overpriced hype job” and “does not reveal bad news.” The market disagreed; Medco’s stock price rose from $6.75 in the following month to more than $30 a year later. In May 1991, a business journal reported that Medco had “found a whole new crew of idiots to buy [its] stock.” Again the market disagreed. All the storm warnings posted before April of the following year were premature. An idiot who bought stock in Medco in January 1991 and sold it in June 1992, at the very bottom of Medeo’s plunge, would still have made a profit of more than 100 percent.

Throughout this period Medco was one of the most shorted stocks on the American Stock Exchange. The judge thought this should have warned investors that Medco was in trouble. Not so. For every short seller — a pessimist about the value of the stock that he’s selling short — there is, on the other side of the transaction, an optimist, who thinks the stock worth more than the short-sale price. Unless the shorts are trading on insider information, all that a large volume of short selling proves is a diversity of opinions about the company’s future — a diversity hardly surprising when the company’s future depends on when the Food and Drug Administration will permit it to sell a new drug. Of course, if there were more pessimists, all wanting to sell short, than there were optimists, the price of a stock would plunge; but the important thing would not be the short selling, but the price plunge, and we made clear in our previous opinion that a price plunge, without more, is not a reasonable basis for suspecting fraud.

The strongest evidence submitted by Medco in support of its motion for summary judgment was a series of articles published in August of 1992 reporting that Fujisawa had sued the principal owner of the company that had sold it LyphoMed. (That suit was decided against Fujisawa in Fujisawa Pharmaceutical Co. v. Kapoor, 936 F.Supp. 455 (N.D.Ill.1996), now pending on appeal to this court.) In part the articles merely repeated what had been reported in June — that Fujisawa had had to recall some batches of Adenocard. But they also reported that the former LyphoMed (now Fujisawa) was having quality-control and regulatory-assurance problems with a number of LyphoMed drugs. Coming only four months after Medco had represented to the investing public that its application for FDA approval of Adenoscan was proceeding smoothly, these reports of trouble, in conjunction with the plunge in Medco’s stock price, created grounds for suspicion that Medco’s representation about being “on track” might have been false. Missing, however, was any indication in the articles published in August of 1992 that Medco had been aware of the problems at Fujisawa-LyphoMed, or aware of their bearing on the prospects for early approval of Adenoscan, when four months earlier it had announced that everything was going smoothly. One of the plaintiffs’ claims of fraud is that Medco *785 did know then. But nothing in the trade press indicated that. An innocently false representation is not actionable under Rule 10b-5. Ernst & Ernst v. Hochfelder, 425 U.S. 185, 193, 96 S.Ct. 1375, 1380-81, 47 L.Ed.2d 668 (1976); Eckstein v. Balcor Film Investors, 8 F.3d 1121, 1131 (7th Cir.1993); Goldberg v. Household Bank, F.S.B., 890 F.2d 965, 967 (7th Cir.1989). When the representation is false for reasons likely to have been within the knowledge of the company when making it, investors upon learning of the falsity should smell the possibility of fraud, but not when the representation concerns conditions internal to a customer, supplier, licensee, licensor, or other outsider to the firm making the representation, conditions that moreover may have arisen after the representation was made.

Another claim of fraud is that back in August of 1991, when Medco had characterized the FDA’s request for additional data as a sign that approval was imminent, it knew that the request signified that there would probably be a substantial delay. Nothing in the trade press would have alerted investors to this fraud either. Taken as a whole, the media reports on Medco revealed widely different assessments of the value of the stock and (what was basically the same thing) the likely date of approval of Adenosean, but did not scatter clues that Medco had been lying about its dealings with the FDA.

But suppose it were true that before September 1992 investors knew enough to suspect fraud. Would the statute of limitations have begun to run then? As an original matter, one might well answer “yes” but then quickly add that the running of the statute of limitations would be tolled (interrupted), as permitted by the doctrine of equitable tolling, until the investors had enough facts in hand to enable them to file a complaint that would comply with the requirements of the Federal Rules of Civil Procedure for pleading fraud. These requirements include pleading the fraud with particularity (a requirement that has been stiffened for securities fraud cases filed after December 22, 1995, see Private Securities Litigation Reform Act of 1995, Pub.L. No. 104-67, § 101(b), § 21D(b)(2), 109 Stat. 737, 747, codified at 15 U.S.C. § 78u-4(b)(2)) and having a reasonable evidentiary basis for all factual allegations in the complaint. Fed.R.Civ.P. 9(b), 11(b)(3); In re Healthcare Compare Corp. Securities Litigation v. Healthcare Compare Corp., 75 F.3d 276, 280-81 (7th Cir.1996); In re Donald J. Trump Casino Securities Litigation, 7 F.3d 357, 373 n. 17 (3d Cir.1993). However, the Supreme Court has held that, given the discovery rule, there is no defense of equitable tolling to the statute of limitations in a Rule 10b-5 case. Lampf Pleva, Lipkind, Prupis & Petigrow v. Gilbertson, 501 U.S. 350, 363-64, 111 S.Ct. 2773, 2782-83, 115 L.Ed.2d 321 (1991). This holding makes it extremely important to decide what exactly it is that the investors must discover (or should have discovered) to start the statute of limitations running. Is mére suspicion discovery? Or, at the other extreme, must the investor have learned (or have been in a position where he should have learned) all the facts he needs in order to file a suit? Language can be found in the ease law to support either formula, and although more cases use the former, see, e.g., LaSalle v. Medco Research, Inc., supra, 54 F.3d at 444, and Whirlpool Financial Corp. v. GN Holdings, Inc., 67 F.3d 605, 609 (7th Cir.1995), than the latter, see Dodds v. Cigna Securities, Inc., supra, 12 F.3d at 350, none of the cases treats the difference between the two formulas as an issue. See also Norris v. Wirtz, 818 F.2d 1329, 1334 (7th Cir.1987), overruled on other grounds in Short v. Belleville Shoe Mfg. Co., 908 F.2d 1385 (7th Cir.1990). Perhaps it made no difference in those cases or the parties had chosen not to make it an issue.

We think it is time to resolve the issue. The most sensible approach, it seems to us, is to adapt the formula that section 13 of the Securities Act of 1933 uses for its one-year statute of limitations for suits complaining of false registration statements: “after the discovery of the untrue statement ... or after such discovery should have been made by the exercise of reasonable diligence.” 15 U.S.C. § 77m. In other words, the plaintiff gets a year after he learned or should have learned the facts that he must know to know that he has a claim. In the case of a suit complaining of a false registration statement, all he *786 has to know is that the statement was untrue; so, as soon as he knows or should know that, the one-year period begins to run. In a fraud ease, he needs to know more: that the defendant has made a representation that was knowingly false. When the plaintiff knows or should know this, the statute of limitations begins to run. This approach is implied in Lampf itself. The Court said not that equitable tolling was inconsistent with the one-year statute of limitations (as it was, the Court held, with the three-year statute of repose), but that it was “unnecessary” because of the discovery rule. 501 U.S. at 363, 111 S.Ct. at 2782. This formulation implies that the rule should be so interpreted as to make equitable tolling unnecessary to protect investors’ interest in having a reasonable, a practical, time within which to sue. It may not have been an accident, therefore, that the Court described the discovery rule as requiring that suit be “commenced within one year after the discovery of the facts constituting the violation.” Id. at 364, 111 S.Ct. at 2782 (emphasis added).

The test is an objective one, as the statutory language makes clear: not whether the plaintiffs did know more than a year before they sued that the defendant had committed fraud, but whether they should have known. Suspicious circumstances, coupled with ease of discovering, without the use of legal process, whether the suspicion is well grounded, may cause the statute of limitations to start to run before the plaintiffs discover the actual fraud, as in Renz v. Bee man, 589 F.2d 735, 751-52 (2d Cir.1978). The defendants in our case, however, have not explained what investors could have done before Medco sued Fujisawa in May of 1993, less than three months before this suit was filed, to obtain the necessary facts. We pressed the defendants’ lawyer on this point at argument, and he suggested that the plaintiffs should have hired a lawyer to investigate, called their broker, or called Medco. These do not strike us as serious suggestions. The lawyer would not be able to subpoena the correspondence between Med-co and the FDA which when it finally came to light in Medco’s suit against Fujisawa established the factual predicate for the claim of fraud. The defendants do not argue that the lawyer could have obtained the documents from the FDA under the Freedom of Information Act, which has an exception for trade secrets and commercial information. 5 U.S.C. § 552(b)(4).

The statute of limitations in securities fraud cases serves, as we have emphasized in other opinions, important public purposes. Tregenza v. Great American Communications Co., 12 F.3d 717, 722 (7th Cir.1993); Short v. Belleville Shoe Mfg. Co., supra, 908 F.2d at 1392. But too much emphasis on the statute of limitations can precipitate premature and groundless suits, as plaintiffs rush to beat the deadline without being able to obtain good evidence of fraud; and the three-year statute of repose gives defendants a definite limit beyond which they needn’t fear being sued. On the record compiled so far, the defendants in this case, who have the burden of proving an affirmative defense, such as that the statute of limitations has run, have failed to show that a reasonably diligent investor would have brought suit before this suit was actually filed.

The defendants’ alternative ground for affirmance is based on a study by a finance expert that compared the price movements in Medco’s stock from November 1990 to June 1992 — the period in which it first rose by 375 percent to its peak value in January of 1992 and then plunged by 61 percent from that value — with the price movements in Medco’s competitors’ stocks. The prices of those stocks had risen and fallen in tandem with Medco’s, and the expert concluded that general market forces, rather than anything special to Medco, specifically the representations in May 1991 and April 1992 that exaggerated the prospects of early approval of Adenoscan, had been responsible for Medco’s decline. If this is right, the fraud caused no harm, and the suit fails. Bastian v. Petren Resources Corp., 892 F.2d 680, 685-86 (7th Cir.1990); Roots Partnership v. Lands’ End, Inc., 965 F.2d 1411, 1419 (7th Cir.1992); Caremark, Inc. v. Coram Healthcare Corp., 113 F.3d 645, 648-49 (7th Cir.1997); Citibank, N.A v. K-H Corp., 968 F.2d 1489, 1495-96 (2d Cir.1992); McGonigle v. Combs, 968 F.2d 810, 821 (9th *787 Cir.1992); see also Private Securities Litigation Reform Act of 1995, § 101(b), § 21D(b)(4), 109 Stat. at 747 (15 U.S.C. § 78u-4(b)(4)) (codifying the judge-made “loss causation” rule). It may or may not be right; but the plaintiffs’ reply brief does not mention the study, even though the defendants rely on it in their brief to provide the factual basis for the argument that the alleged fraud did not cause the plaintiffs’ loss.

Failure to contest a point is not necessarily a waiver, but it is a risky tactic, and sometimes fatal. See Hardy v. City Optical Inc., 39 F.3d 765, 771 (7th Cir.1994); Singletary v. Continental Ill. Nat'l Bank & Trust Co., 9 F.3d 1236, 1240 (7th Cir.1993). Here it leaves us without any basis for questioning the soundness of the defendants’ study. It is true that the plaintiffs presented some contrary evidence in the district court. But they do not mention that evidence in this court. By their silence the plaintiffs imply either that they think evidence on causation irrelevant — which would bespeak a fundamental misunderstanding of the law governing private damages actions for securities fraud — or that they have lost faith in their own study but hope to find better evidence if the case is remanded. They could have said simply that given their own study, the defendants’ evidence on causation is thrown sufficiently into doubt to warrant a trial. But they have not said that. We conclude that the suit was properly dismissed, although not on the basis of the statute of limitations.

Affirmed.