delivered the opinion of the court:
*324This patent case comes to us on the Government’s exceptions to Trial Judge Browne’s decision in favor of plaintiff McDonnell Douglas Corporation. Plaintiff seeks compensation under 28 U.S.C. § 1498(a) (1976) for the allegedly unauthorized purchase of an antitank missile system said to be covered by plaintiffs patent, and also under 35 U.S.C. § 183 (1976) for damages resulting from the imposition of a secrecy order on the missile design. We reverse the decision of the trial judge on the former issue and remand for further proceedings on the latter.
In 1959, McDonnell Douglas Corporation (McDonnell) began work on an anti-target missile system. By 1962, McDonnell had successfully developed, built, and tested the TOW SIDEKICK antitank missile. Soon thereafter, when the Army expressed the need for a Medium Antitank/Assault Weapon (MAW), McDonnell responded by proposing a scaled down version of the TOW SIDEKICK utilizing similar guidance logic. Subsequently, James M. Tucker, a McDonnell employee, conceived a new type of guidance system for the MAW missile. This new guidance system and the scaled down version of the TOW SIDEKICK missile were the subject of Patent Application Serial Number 346,165 (the Tucker Patent Application) which was filed on February 20, 1964, and eventually matured into U.S. Patent No. 3,868,883 (the Tucker or ’883 patent). The patent had not yet been granted, however, when, on August 31, 1964, McDonnell entered into an experimental development contract, Contract No. DA-01-021-AMC-11350(Z) (the 11350(Z) contract), with the Army to develop such an MAW missile. At the time the 11350(Z) contract was made, the new, smaller McDonnell missile had not been constructed or flight tested, although the guidance system and other subsystems had been tested by computer simulation. McDonnell successfully completed the experimental development program under the 11350(Z) contract, but, in the end, the Army decided to buy its MAW missiles from McDonnell’s competitors, Raytheon Corporation and Kolls-man Instrument Company.1
*325Plaintiff asserts that the defendant’s procurement of missile systems (called Dragon) from these other companies infringed McDonnell’s rights under the Tucker patent. Whether the purchase infringed the ’883 patent depends on two questions: does the Government have rights to use the patent, and, if not, does the purchased missile infringe a claim of the ’883 patent. We hold that the Government had rights in the ’883 patent sufficient to allow it to make the purchases from the competitor firms, and therefore we need not decide infringement. See Leesona Corp. v. United States, 208 Ct. Cl. 871, 887 note 9, 894, 530 F.2d 896, 906 note 9, 910 (1976), Mine Safety Appliances Co. v. United States, 176 Ct. Cl. 777, 781, 364 F.2d 385, 387 (1966).
I
The basic problem is whether the Government had a right to use or purchase the allegedly infringing articles. Defendant contends that that right came to it from the Subject Invention clause of the 11350(Z) contract because the invention was completed (by reduction to practice) under that agreement. Plaintiff answers, first, that the Tucker invention was wholly excluded by the parties from the Subject Invention clause, and, second, that the invention was reduced to practice before the consummation of the Government’s 11350(Z) contract and accordingly was outside the scope of the Subject Invention clause. In this part of our opinion we consider plaintiffs first argument.
The Subject Invention provision reads: "Except as provided in (e) and (h) of this clause, the Contractor agrees to grant the Government all right, title and interest in and to each Subject Invention (made by the Contractor), subject to the reservation of a nonexclusive and royalty-free license to the Contractor.” Pt. II, art. D, cl. 22 of the contract as amended by the May 11,1965 modification incorporating 32 C.F.R. § 9.107-5 (1965) (adopted July 21, 1964).2 The con*326tract and the regulations then define a Subject Invention as "any invention or discovery, whether or not patentable, conceived or first actually reduced to practice in the course of or under this contract.” Id.3
McDonnell’s first position (which was accepted by the trial judge) is that the parties agreed wholly to exclude the Tucker invention from the Subject Invention clause. For us, the overriding factor resolving this contractual question is that, under the regulations in effect when the contract was made on August 31, 1964 (unlike the regulations they replaced), the contracting officer had no authority to give, by himself, contractual exemptions from the patent rights clause. Compare 32 C.F.R. §9.107-2(a) (1961) with 32 C.F.R. §9.107 (1965). (The July 21, 1964 amendments eliminated the possibility of such an exclusion for a particular invention from the patent rights clause.) Any attempt to do so would constitute a deviation from the regulations. 32 C.F.R. § 1.109-1 (1964) (definition of deviation). Deviations were allowed only if authorized in accordance with departmental procedures and if written notice of the deviation was furnished to the Assistant Secretary of Defense and the Assistant Secretary of the Army. 32 C.F.R. § 1.109-2 (1964).4 Plaintiff has produced nothing to indicate that such authorization was granted or that proper notice was furnished, and we can take it that they were not given. In this case, therefore, any deviation from the prescribed wording was a violation of the regulation and beyond the authority of the contracting officer. Actions taken by that officer beyond his authority are, of course, not binding on the Government. E.g., Federal Crop Ins. Corp. v. Merrill, 332 U.S. 380 (1947). It is in the light of those cardinal rules that the plaintiffs contentions must be assessed.
The Subject Invention provision contains no exception for the Tucker invention, but McDonnell says that clause 23 of *327the contract, "Rights in Technical Data”, modifies the unqualified patent rights clause (including the Subject Invention provision). That data rights article limits the Government’s right to disclose much technical data submitted by McDonnell, including that in the Tucker patent application.5 Because the clause in this contract specifically mentions the ’883 patent, plaintiff urges that the Tucker patent is thereby excluded from the coverage of the patent rights clause. The data rights clause itself, however, belies that conclusion. Paragraph (e), "Relation to Patents”, of that clause states: "Nothing contained in this clause shall imply a license to the Government under any patent or be construed as affecting the scope of any license or other right otherwise granted to the Government under any patent.” (Emphasis added). This paragraph shows that the data rights clause is intended to have no effect on patent rights, whether license or title, granted to the Government by the 11350(Z) (or any other) contract. In addition, rights to technical data (the sole subject covered by the technical data clause) normally differ from patent rights (cf. e.g., Kewanee Oil Co. v. Bicron Corp., 416 U.S. 470 (1974)), and it would have been illegal, as we have pointed out, for the contracting officer to use this separate means as a covert and unacknowledged effort to exclude the ’883 patent from the Government’s "subject invention” rights. The terms and policy of applicable procurement regulations are to be followed, not thwarted. See Chris Berg, Inc. v. United States, 192 Ct. Cl. 176, 182-83, 426 F.2d 314, 317 (1970); Firestone Tire & Rubber Co. v. United States, 195 Ct. Cl. 21, 34-35, 444 F.2d 547, 553-54 (1971); ITT Arctic Services, Inc. v. United States, 207 Ct. Cl. 743, 763, 524 F.2d 680, 690 (1975). A corollary is that contract provisions are to be read, if fairly possible, to accord with those principles.
McDonnell then refers to subsequent contracts. In three later agreements the following language, or similar wording, appears: "Nothing contained hereinabove or elsewhere in this contract shall imply a license to the Government *328under a patent issuing from the [Tucker Patent Application] or be construed as affecting the scope of any license or other right granted to the Government under this or any other Government Contract.” Plaintiff, seizing upon the first half of this sentence, maintains that it proves that the Government had no right or interest in the ’883 patent. The conclusion of the sentence indicates otherwise. As a whole, the sentence does not act as a waiver, as the trial judge found it did, of the Government’s pre-existing claims or rights, but, rather, makes clear that the later contracts are to have no effect on any and all pre-existing rights. Here, too, it would have been inadmissible for the contracting officer to flout the regulation by seeking (without proper authorization and notification) to remove the ’883 patent from the Subject Invention provision.
Another of plaintiffs points is that the list of Subject Inventions submitted by McDonnell, as required by the patent rights clause, did not include the ’883 patent. From the defendant’s failure to challenge the omission, it is argued that the Tucker patent could not be a "Subject Invention.” Nothing in the contract supports that conclusion. The purpose of the reporting requirement is not to limit the Government’s rights to inventions but to apprise the United States of the inventions in which it has an interest. To the extent that the plaintiff is arguing estoppel, that issue has already been expressly decided against it. This court, in its previous decision on cross motions for summary judgment, held that the Government had no obligation to assert its right in the ’883 patent at any time prior to the institution of this suit. McDonnell Douglas Corp. v. United States, 222 Ct. Cl. 570 (1980).
It may be that McDonnell always wanted and intended to retain full patent rights to the Tucker invention and the ’883 patent, but there is no adequate showing that the defendant ever agreed with that position — and the contracting officer could not have done so without complying (as he did not and may not have wanted to try to do) with the controlling regulation. The most that could be (and was) done was to protect plaintiffs rights in technical data. Plaintiff thus agreed to the full Subject Invention clause *329without ever obtaining a proper exception for the ’883 invention.6 None of the provisions in the 11350(Z) contract or the other government agreements should or can be read to state or imply such an exception. The United States therefore had sufficient rights to acquire and use missiles reading on the ’883 patent so long as that invention qualifies as a "Subject Invention” under the contract.
II.
The next issue is whether the Tucker invention was a "Subject Invention” under the 11350(Z) contract. Such an invention is one "conceived or first actually reduced to practice in the course of or under” the 11350(Z) contract. See Part I, supra, text at note 3. Because it is agreed that the ’883 missile was conceived before the date of contracting, and was, at the latest, operational by the end of the contract, the focus of inquiry narrows to whether it was reduced to practice before August 31, 1964 (the date of the contract). McDonnell, both because it is the plaintiff and because it is the party asserting the prior reduction to practice, has the burden of showing that the device had been reduced to practice before that critical date. See Lockheed Aircraft Corp. v. United States, 213 Ct. Cl. 395, 405, 553 F.2d 69, 74 (1977); cf., Coffin v. Ogden, 85 U.S. (18 Wall.) 120, 124 (1873) (burden of proof rests on party asserting earlier invention); Perry v. United States, 112 Ct. Cl. 1, 33-34, 76 F. Supp. 503, 507 (1948) (plaintiff asserting reduction prior to an anticipatory disclosure shoulders strict burden).
Reduction to practice occurs when it is established that the invention will perform its intended function beyond a probability of failure. General Electric Co. v. United States, 228 Ct. Cl. 192, 204, 654 F. 2d 55,62 (1981); Eastern Rotorcraft Corp. v. United States, 181 Ct. Cl. 299, 302-03, 384 F.2d 429, 431 (1967). This usually requires testing of the invention. Id. The question is whether the tests done by McDonnell prior to August 31, 1964 "were sufficiently comprehensive to demonstrate the workability of the *330device;” Bendix Corp. v. United States, 220 Ct. Cl. 507, 517, 600 F.2d 1364, 1370 (1979).
Plaintiff does not argue that it met this standard by an actual firing of a physical model of the missile. Indeed, such a physical embodiment of the claims of the Tucker patent did not exist until some time in 1965. Rather, McDonnell insists that the invention was reduced to practice by computer simulation. The theoretical issue of whether an invention can ever be reduced to practice by computer simulations has been argued vigorously by both parties. But the salient facts in this case obviate the need to decide that novel and difficult problem. If there can ever be a valid reduction to practice via computer simulation, that conclusion could not be reached when actual physical testing has demonstrated that prior computer simulation was inadequate. That is this case — the physical testing under the 11350(Z) contract showed that the earlier computer simulation was insufficient to prove the proper workability of the device or that it would probably perform its function. As we shall now detail, the physical testing revealed significant flaws in the plaintiffs design, making the device (as passed by computer simulation) incapable of meeting importánt elements of the patent claims now involved.
The trial judge found that claims one and six of the ’883 patent had been infringed by the Government. Plaintiff and defendant have treated those claims as representative of the claims allegedly infringed, and so shall we. Claim one comprises a "means on the [missile] responsive to said commands to control the frequency of excitation of succeeding groups of thruster elements * * *” and claim- six similarly includes a "means on the missile for sequentially energizing the thrust producing elements at a frequency under control of the first command signals * * *”. Thus, each of the claims specifies that there shall be some means to control precisely the sequence of thruster firings. However, the means that had been designed and tested by simulation before August 31,1964 suffered some important design defects.
First, because of a design oversight, under certain conditions the control system electronics unit (CSE) would not reset properly after the thrusters fired, and another pair of *331thrusters would be excited. Two or more firings could thus occur when only one should have taken pláce. This defect, in one test, deflected the missile seven feet from its intended path. Although the missile could recover from such a misfire, provided that another double firing did not occur, the recovery would not be completed until the missile had hurtled more than 100 meters past the point of misfire. If multiple misfires occurred, or if a single misfire occurred shortly before the missile was to hit the target, it is not unlikely that the missile would miss the target entirely. The seriousness of this defect was discovered only during physical testing.
Another control system (CSE) design defect was also first discovered in tests conducted in the course of the contract— the CSE would not operate under high humidity conditions. Tests which fail to simulate the varying and multiple conditions of the invention’s intended environment will not serve to prove the operability, stability and reliability of the invention for practical use. Technical Development Corp. v. United States, 202 Ct. Cl. 237, 308-10 (1973), cert. denied, 416 U.S. 983 (1974); see Leesona Corp. v. United States, 208 Ct. Cl. 871, 895, 530 F.2d 896, 910 (1976). McDonnell’s precontract tests did not test the guidance system under high humidity conditions (hardly an improbable circumstance in the conditions for use of a missile designed to be launched against a tank or comparable moving target) and must be considered inadequate when the physical tests revealed non-operability under those conditions.
Another example illustrates nicely the inadequacy of the computer simulation, and the necessity, in this case, for physical testing. The modification implemented to correct the double firing defect caused a new fault; the CSE became too sensitive to radio frequency noise. Further design changes were necessitated by the discovery of that problem. The original computer testing program, because it was unable to detect the initial defect, simulated a device that did not incorporate a solution to that flaw, and was consequently unable to detect or predict problems arising from the retrofitted solution.7
*332These examples make clear that the means for properly guiding the missile were not yet shown to be workable on August 31, 1964. Modifications were required before the ’883 invention would work beyond a probability of failure. The necessary changes were not mere "minor adjustments”, as in Bendix (220 Ct. Cl. 507, 518, 600 F.2d 1364, 1371 (1979)) but, rather, were of "critical importance to achieving the objectives of the invention,” as in Leesona (208 Ct. Cl. at 895, 530 F.2d at 910). The physical tests proved that the computer approved device, designed to be launched by a person against a still or a moving target, failed in actual practice (in several significant circumstances) to perform in the manner intended. In short, the physical tests demonstrated that the computer-tested device might or might not work as it was supposed to do.8
Our conclusion that the physical testing under the 11350(Z) contract, by indicating the insufficiency of the prior computer tests, was the first actual reduction to practice accords both with the objectives of the concept of reduction-to-practice and with the purposes of the Subject Invention clause. The aim of the reduction-to-practice requirement is to ensure that the new device is sufficiently *333"complete and capable of working”, Coffin v. Ogden, 85 U.S. (18 Wall.) 120, 125 (1873), that it may enter into the stream of commerce and of use. That kind of disclosure benefits the public, and the benefit justifies granting a right of exclusive production and control to the inventor. But if the product is not sufficiently developed to enter the marketplace, the public does not benefit and the grant of a monopoly cannot be justified. See Coffin v. Ogden, 85 U.S. (18 Wall.) 120, 125 (1873); cf., Gayler v. Wilder, 51 U.S. (10 How.) 477, 497 (1850). Here, the physical testing was needed to show the capability of actual working; that status was not yet proved by the computer tests alone. The physical testing was therefore the first actual reduction to practice.
One major aim of the Subject Invention provision is to protect the Government from having to pay for using an invention which it significantly helped to develop. See Mine Safety Appliances Co. v. United States, 176 Ct. Cl. 777, 789, 364 F.2d 385, 392 (1966); Ordnance Eng’r Corp. v. United States, 68 Ct. Cl. 301, 353, (1929), cert. denied, 302 U.S. 708 (1937); Technitrol, Inc. v. United States, 194 Ct. Cl. 596, 613, 440 F.2d 1362, 1372 (1971); Technical Development Corp. v. United States, 202 Ct. Cl. 237, 255-56 (1973), cert. denied, 416 U.S. 983 (1974). That type of significant aid was plainly given here by the 11350(Z) agreement. While McDonnell seems to have spent less than $600,000 on its development of the device before that contract, the Government expended more than $3 million (under the contract) in building, testing, and modifying the missile. Most important, that post-contract development led to essential changes in the device and first proved it workable. As the Subject Invention clause contemplated, the United States has and should have the right to procure and use the Tucker invention without payment.
III.
McDonnell also makes a claim, under 35 U.S.C. § 183 (1976), for damages caused by a secrecy order imposed by the Government on the Tucker Patent Application. Section 183 provides for damages to be paid to patent holders in two situations: when the Government wrongfully uses the *334patented device during the period of secrecy, and when the secrecy order itself causes damages. See generally Constant v. United States, 223 Ct. Cl. 148, 617 F.2d 239 (1980). To the extent that the §183 claim depends upon issues also present in plaintiffs current infringement claim, it is obviously vitiated by our decision. The plaintiff cannot recover on a theory of infringement or improper use, because we have found that the defendant had a right to buy and use the invention. Plaintiffs only remaining claim under § 183, then, is for damages caused by the secrecy order itself. McDonnell’s petition does not allege any potential commercial use of the missile. However, if such non-government use can be established, as required by § 183, damages flowing from the secrecy order may conceivably be recovered. We remand for possible further proceedings on the secrecy claim.
IV.
Accordingly, we hold that plaintiff has no valid claim against defendant for patent infringement. Plaintiff is free, on remand, to assert a § 183 claim based on damage to plaintiffs ability to exploit commercially the ’883 patent (if such damages were caused by the secrecy order). Plaintiff may also amend its complaint to assert a claim for breach of the data rights clause. See note 5, supra. We, of course, take no position with respect to the validity of those claims.
CONCLUSION OF LAW
The court concludes as a matter of law that the plaintiff is not entitled to recover under 28 U.S.C. § 1498(a) (1976) and that claim of its petition must be dismissed. Leave is granted plaintiff to amend its petition to assert a claim for breach of the data rights clause in the 11350(Z) contract; the case is remanded for further proceedings on that issue. The trial judge is also free, on remand, to consider any claims based on 35 U.S.C. § 183 (1976) not precluded by our conclusion that the Government is fully free to use the ’883 patent and the Tucker invention without cost.
The facts necessary to this decision are stated in this opinion.
The exceptions do not affect the result in this case. The first exception, paragraph (e), requires certain filings to be made in cases where greater rights are given to the contractor pursuant to paragraph (h), but does not affect the scope of the rights that can be granted. The other exception, paragraph (h), allows greater rights to the contractor, but is expressly limited by the provisions of paragraph (i) providing that "such greater rights shall, as a minimum, be subject to an irrevocable, nonexclusive *326and royalty-free license to practice and have practiced the Invention throughout the world for Governmental purposes.” At most, therefore, the exceptions reduce the rights granted the United States to a royalty-free license — which would be enough to immunize defendant for using and purchasing the accused articles here.
In this connection, defendant’s claim is that the Tucker invention was reduced to practice under the contract, the plaintiffs that it had previously been reduced to practice. See Part II, infra.
McDonnell was informed (prior to the making of the contract) of the relevant changes in the military’s patent regulations.
Plaintiff has not asserted a claim based on an alleged breach of this contractual provision; it argues here only that the provision modifies the patent rights clauses. Because it appears that the Government may possibly have violated the data rights clause, we allow plaintiff to amend its petition to assert such a claim. See Part IV, infra.
There has been no attempt to allege or show fraud or misrepresentation on the part of the Government.
Defendant has also shown that the physical testing under the 11350(Z) contract revealed that the prior computer simulation had utilized faulty data on gunner error *332in aiming and guiding the missile. (The trial judge found (see his finding 84), and plaintiff accepted the finding, that the device was intended to be carried by a single person who operated the guidance system leading the missile to its target). If, as seems the case, this understatement of gunner error was substantial and relatively widespread, leading to target failures or misshits, this defect likewise proves the inadequacy of the computer simulations.
Plaintiff’s own test report identifies still another shortcoming of the simulations done prior to August 31, 1964. In describing the system integration tests (SIT), the report states: "The purpose of the SIT was to evaluate the integrated performance of all subsystems under simulated flight conditions (with the exception of missile displacement). The components and subsystems had been evaluated individually and in pairs, as indicated above, however, in the SIT all were combined for the first time.” Before the SIT "the proof of compatibility of all equipment remained to be proven. In the SIT the following items were evaluated for the first time:
1) Operation of tracker/guidance wire link/CSE loop,
2) Firing of side thrusters using gyro reference,
3) Effect of side thruster firings on roll rate,
4) Effect of side thruster firings on tracker operation,
5) Complete telemetry and ground instrumentation operation, and
6) All missile equipment mounted in missile structure.”
The SIT took place in 1965. Especially when the invention is so complex, it is difficult to maintain that it was reduced to practice in 1964 when the compatability of its components remained in question until early 1965.
It is not unimportant that eight out of the first nine test flights failed to hit the target. This was only an 11% success rate. Perhaps some failures were due to factors having nothing to do with design or inventive defects, but it seems improbable that all or even most were.