This seems to be a point that many people are confused by. If your email is being correctly classified with the corpus that you have set up currently, there is no reason to add more data to the corpus. POPFile only modifies the corpus when it misclassifies an email and it does so when you reclassify the message itself. The assumption is that the data in the corpus is all that is needed for classification. If POPFile makes a mistake and you choose to reclassify a message, it adds more data to the corpus so that it will not make that same mistake again. POPFile does not need a lot of information in its corpus to be extremely effective.