Honolulu and Fort Lauderdale, Fla., tied for third-best in the tests, both failing to detect the fake weapons 10 percent of the time. Miami fared best at 6 percent, followed by Newark, N.J., at 9 percent.

The tests, the first since the security agency began overseeing checkpoint screening in February, were done by agents who were instructed to do little to try to conceal the items as they passed through screening checkpoints, memos about the tests show.

Acting Honolulu airport administrator Roy Sakata today said he has not seen the report, the results of which are supposed to be confidential.

"We always try to minimize," he said. "We are always striving to get it down to catching everything that goes through there."

Overall, screeners missed simulated weapons in 24 percent of the tests. At three major airports ­ in Cincinnati, Jacksonville and Las Vegas ­ screeners failed to detect potentially dangerous items in at least half the tests. At a fourth, Los Angeles International Airport, the results weren't much better. The failure rate there was 41 percent. Screeners repeatedly failed to find stainless-steel test pieces that set off metal detectors as guns might. Screeners also had trouble spotting simulated bombs.

"A 41 percent failure rate is just pathetic," says Jack Plaxe, an aviation security consultant. "There has to be problems with the people or their training."

Nationwide, screeners often failed to find simulated weapons on agents after metal-detector alarms sounded. In 178 tries, screeners failed to find potentially dangerous items on agents in a third of the tests.

At some of the 32 airports, agents conducted only a handful of tests. At the 12 airports where at least a dozen tests were conducted, the failure rate was 29 percent.

The documents detailing the results are part of a series of undercover tests that are set to conclude today. The screeners who were tested had been trained by security companies that used to work for the airlines and which the TSA now oversees. Tens of thousands of them likely will be hired by the government by November, when screeners will become federal employees. The TSA plans to deploy about 45,000 screeners by then.

Of the 387 tests, 209 involved screeners operating X-ray machines. The failure rate was 16 percent. The other tests assessed whether screeners detected objects that set off metal-detector alarms.

The results raise questions about whether screening has improved since the TSA took responsibility for overseeing airport checkpoints.

In tests completed earlier this year before the federal takeover, investigators with the Transportation Department's independent watchdog, the inspector general, found failure rates of nearly 50 percent at 32 airports that they tested.

But the manner in which those tests were done differed from the TSA's approach. TSA agents were instructed to pack bags containing the simulated weapons "consistent with how a typical passenger in air transportation might pack a bag."