No you aren't missing anything:

* managers that underperform tend to get fired

* managers that outperform tend to get hired

* but if you track the performance of fired managers, you find that they tend to outperform their benchmarks

* and recently hired managers tend to underperform their benchmarks.

This would all be consistent with a large noise component to active returns. If these cycles of performance ups and downs coincide with manager reviews then you get the perfect recipe for destroying active returns