> Empirically, CTVP attains very good detection rates with reliable false positives

A novel use of the word "reliable"? Jokes aside, either they mean the FPR as the opposite of what you'd expect, the table is not representative of their approach, or they're just... really optimistic?