> "so there's a real emphasis on only hiring perfect "A"s. Which means that you'll miss a few As and lots of B+ people."

It's actually worse than that. When you exclude huge swaths of a sample like that, you have no context for the true scale.

So while you can hire the best of the sample you see, while understanding you may miss other people nearly that good, you may also be missing people who are better (perhaps wildly so) than even the best of those you chose.

Consider the case of competitive sports during times of segregation.