That's a myth. The real problem is that the average teacher (relative performance) is so far away from being close to maximally performant (as proved in studies which rigourously control teacher behaviour) that the impact of variation about that mean is swamped by environmental factors. People conclude that since swapping teachers doesn't make an appreciable difference, teacher performance can't make a difference. When in reality the problem is that teaching is uniformly bad (compared to maximal possible performance, which is very difficult. this is why the answer is not to look for better teachers but to look for better, uniformly implementable, curriculums)