I'm sorry but I don't see how this addresses my point? If they are checking 20 bands for results with p < 0.05, the likelihood is that one will show up, just from fluctuations from background.