Do not ignore empirical evidence - reality will always crush theory (in the collequial sense).

And empirically, we endlessly observe very nervous people at interviews that do not exhibit nervousness in meetings (after getting the job).

We know that people behave differently in interviews; I suggest using that behavior to predict behavior in non-interview situations is noisy at best, and probably entirely random.