I think the point is that the underlying metrics have a degree of randomness which we need to recognize. Instead of just deciding what is good enough to make it at Harvard and have that be the cutoff, you could certainly set the cutoff higher but still have randomness after that point. Don’t introduce arbitrary sorting when that is really not possible.

I TA’d a class where we used a grading metric sort of like this. It turns out that it is a lot easier to grade assignments by grouping them into 5 buckets than it is to give an extremely precise point value to everything.