One of the weird things about evaluating teacher quality through "value-added" testing data is that most people would say that the results actually underestimate the true difference in value of a great teacher over a lousy one. Think of a fantastic teach you or your child has had; now think of a bad one. Is the first only three times better than the second? I think not.
It is only when you artificially narrow the definition of good teaching that the impact of teacher quality seems so small.