Schneier is one of the great popularizers of the idea that there can be no security in obscurity -- how can Hawley know that his s33kr1t back-end for preventing moisture bombs and evil shoe-wearing aviation threats works unless it's subject to public scrutiny?
If you don't publish your findings, you're not doing science, you're doing alchemy, and every alchemist had to discover for himself, the hard way, that drinking mercury was a bad idea.
So without getting into specifics on the test results, of course there are times that our evaluations can generate high failure rate numbers on specific scenarios. Overall, though, our ability to detect bomb components is vastly improved and it will keep getting better. (Older scores you may have seen may be "feel good" numbers based on old, easy tests. Don't go for the sound-bite; today's TSOs are light-years ahead of even where they were two years ago.)Link