It isn't. Whether you're forced to retire should be calibrated to the individual's capabilities, not based on a statistical calculation.

Assessments of individuals can be gamed. Hard objective limits cannot.

It's really difficult to game timed and proctored aptitude tests. We've refined these systems pretty well over time. I can't imagine how anyone would game a test like, say, the LSAT unless they had the answers in advance or fraudulently substituted someone to take the test for them. And both cases are pretty easy to detect nowadays.