Comment by lifthrasiir

These metrics necessarily have to underestimate programmer skills because those are not directly controllable. If there is any sort of rigor in these metrics (i.e. I don't know if COCOMO is one of them) they will probably assume, say, a mundane programmer whose performance is worse than 90/95/99% of all other programmers.