I am surprised that a normal distribution was used for the distribution of "leadership skills". I imagine this could be similar to income distribution which has fat tails. I wonder if the result would change if such a distribution was used.
If you relax that assumption even slightly (thus you do not reliably recognize relative abilities, but you tend to prefer the better leader), then the conclusion will change dramatically.
That is why this is a toy model. And people should not give it much weight.