This is a well-known problem with self-assessments in psychology in general. It isn't that hard for most people to choose to use a persona to provide answers to some quiz, even one quite divergent from their "real" personality (skipping the discussion on what that even is), and for some people with certain conditions it's almost impossible for them not to do this (e.g., the "Cluster B" disorders). Getting past this problem is the major task of creating a standardized assessment that has any value at all. You obviously can't just ask "Hey, are you a homicidal maniac who is unfit to stand trial?" but you also can't ask questions that are really quite obviously just that question in disguise like "If someone flips you off in public, do you think you're justified in immediately murdering them?"... and I exaggerate for effect, but the principle holds true.

I'd expect even in the case of this quiz if you told people to affect certain personas you'd find the results statistically-significantly shifted, even if the people involved couldn't tell you any rational reason why they changed their answers based on the persona they affected.