I have only second-hand knowledge, but understand that the software for TSA scans includes randomly placed false positive images to keep the screeners from becoming complacent. [I do know the fellow who consulted for the TSA to determine the most efficient false positive rate, but he would not disclose it to me, nor the true positive rate. General learning theory suggests vigilance is highest with an error rate of about 20%, providing a first-order guess.] From a TSA staffer's perspective who has the knowledge that there are false positives inserted, I can imagine it is easier to re-screen an item rather than to search the bag when there is a positive finding, even though from the perspective of passenger throughput, it may be less efficient to do so.