One could construct tests where this isn’t a given. Create an exam with one really easy question and one really hard question. You’ll end up with a huge spike in the center and two tiny tails. The top quintile of correct answers (<=1.6 correct answers) will be vanishingly small.