That may be the case, but simply running the same code review twice with the same model may give very different results as well. Impossible to know with the information we have.