Teaching requires a set of distinct skills which are different from researching. Naturally not everyone is equally gifted or developed in all areas.

One of the most valuable skills in life is the ability to adapt your mental models and present ideas at an audience-appropriate level. It is useful in teaching; it is also useful in leadership, sales, design...

As an exemplar, I submit Akamai CTO and MIT professor Tom Leighton.

https://www.youtube.com/watch?v=L3LMbpZIKhQ&list=PLB7540DEDD...

This is all true, but learning to teach takes time and one has less time for research.

Too much presenting moves one into extrovert mode where one has many "ideas", but no proofs or execution.

I rarely see a top presenter actually doing anything.

Feynman really is an exception, and there are rumors that his classes were too good -- the students would think they understood everything, but actually did not.

This is also what I experienced when reading his books (which are superbly written).

Sure, Feynmann was pretty special. But the title of the article is "Those who can Do, Can't Teach", which is clearly false given the presence of at least one counterexample. It's possible to make any statement "true" if you don't accept the counterexamples because they are not "statistically representative". :)