I don't know what a false positive is.
Ah, it's a classification term. Basically, a false positive is a principle that is sound by itself, but offers the potential for misleading or incorrect results. To give an easier example, when I worked in Data Security, let's say a policy was created to search for a certain code term, and every time an email was sent that contained a certain term, this email would be flagged as a breach of data. Now let's say the term was the letters "FTH". For the purpose of this example, imagine FTH indicates data that should only be sent privately, not via email.
Now, imagine there is an email that flags up as violating this policy. However, in the email message, someone types the sentence "I really wasn't sure of the correct wording to use". By itself, that sentence is harmless and shouldn't have been caught by that security policy. But if you look closer, the words "of the" are together. OFTHE = FTH. Because the letters FTH are next to each other, the policy flags it up as a breach, thinking a FTH-classified document is being leaked.
This would duly be marked as false positive. Because while it has caught something that matches the policy, the logic behind the policy isn't tight enough, so something that is harmless gets caught.
In the context of Mafia, a false positive scum tell would be a tell that should indicate scummy behaviour, but can also come from Town. Tunneling is a good example.
Long explanation there, but I hope it made sense.