So as long as we manipulate the premise we can control the outcome of the study? Most inmates do not identify themselves as criminals (they're all innocent you know), but their behavior reveals the truth. Behavioral studies should at least have consistency when analyzing deeds. It is what it is.
You're right -- there's no disputing that.
However, the manipulations of these studies aren't looking at behavior. They're looking to score political points by mixing behavior (35% of molestations) and self-identity (2% is homosexual).
More people behave homosexually than identify as "homosexual".
Or, to continue your example, there are more people committing jailable offences than there are people who are "inmates".