People don't talk about this issue anywhere near enough. You can't just do internally ordered evaluations, because someone will eventually compare them to something else and harm everyone who got a harsh 'scale'.
Hence the inability of lone professors or colleges to fight grade inflation; if your students are competing with inflated GPAs from another school (or an expectation that all GPAs are inflated), the incentive is entirely against you.
I suppose this would be the upside of stack-rank if you did it right; a purely relative ranking system can't get poorly translated between groups. But of course, that also means it can't be used to properly distribute benefits between them.