A General Accounting Office report urges the Defense Department to hold back a planned $5 billion program for a controversial advanced Air Force missile called Maverick, saying that five years of operational testing have failed to show that the tank-killing weapon "can be used effectively by U.S. military personnel in combat."

Even under "very favorable test conditions," the missile had only "limited success," the report said. "On the other hand, we find that it is not known whether the . . . Maverick can work well under less-than-favorable test conditions, as may happen in combat."

As an example of "relatively benign test conditions," the report said that at Fort Riley, Kan., last year, the test pilots were briefed in advance and "knew what to look for in the test situation despite the fact that this information would probably not be available in a combat situation."

The pilots "flew in a small and familiar target area that had many unique visual and thermal cues," such as burning hulks that they knew were "enemy" because "friendly" equipment wasn't provided, the report said.

The report stems from a review of the Pentagon's operational weapon testing system, requested by Sen. David H. Pryor (D-Ark.). Pryor wants Congress to create an independent Pentagon office of operational testing and evaluation. The GAO picked Maverick as a case history of weaknesses in current weapon testing, and on June 25 sent the report in classified form to Defense Secretary Caspar W. Weinberger. Pryor released unclassified portions Friday.

A Pentagon panel is scheduled to decide in August whether to order production of 200 advanced Mavericks as the possible first step toward buying 61,000 of the missiles from Hughes Aircraft Co. The GAO report warned against buying the 200 missiles because "even limited procurement requires more evidence of success in testing and evaluation than is currently available."

The Pentagon said Friday that it and the Air Force will have no comment "at this time."

In an interview for a series of Maverick articles published in February by The Washington Post, however, the panel chairman, Defense Undersecretary Richard D. DeLauer said: "You know, it's a go-ahead now, there's no question about it, the question is at what pace . . . ." Similarly, Maj. Gen. James H. Marshall, the Air Force officer in charge of Maverick's development, said: "Based on what we see in the tests, we are confident that it's going to work . . . yes, sir. Absolutely."

The advanced Maverick uses an infrared device to sense temperature contrasts between an object and its hotter or colder immediate surroundings, such as a hot tank exhaust system and the vehicle itself. It shows the contrasts on a five-inch-square cockpit screen, and is supposed to guide the missile to a target. Darkness doesn't affect the heat-seeking sensor, so it was promoted for its capability at night and in "adverse weather." In March, the GAO noted, the quoted phrase was changed to " 'limited adverse weather' without explanation or definition."

The advanced Maverick, which remains in development eight years after the initial procurement for testing, is supposed to succeed the original, daytime-only version, which uses a television camera to detect light contrasts. For the Air Force and the Pentagon, the compelling argument for both the old and new Mavericks is that it lets a tactical aircraft "stand off" at a distance from a target when making a launch.

The unclassified GAO report said, however, that for the missile's primary mission, supporting and protecting ground troops, "It may be questioned whether the IR infrared Maverick can be successfully employed at its purported stand-off ranges." The report added that "the purported advantages of the IR Maverick may be more apparent than real."

The report also questions whether pilots relying on the missile can find the target area "in the day, at night, and in adverse weather"; another is whether they can distinguish a friendly tank from an enemy tank. The report also said that testing up to now has not shown that the missile can't be deceived and defeated either by objects such as sun-warmed rocks or burning battlefield wreckage, or by deliberate countermeasures, such as the firing of a tank cannon that can cause the loss of a "lock" on a target.

The Air Force says that the Maverick will be capable by itself of recognizing a valid target. But the report says the Air Force has been trying to develop a recognition system to help the missile. The system, called LANTIRN, now costs $1.8 billion, and is in technical and financial trouble.