I imagine with many tests, even at 0.1% eventually you'll get some false positive due to human error, or other things going wrong