...what do you do when most kids are getting the same top grades on tests?
Give harder tests and adjust grades upwards accordingly. If everybody is blowing out your measurement scale, your scale needs to be higher.This is only true if the purpose of your measurement scale is to partition scores into a normal (or similar) distribution. This is not always the case - sometimes the purpose of a rank like this is to measure something absolute, rather than relative.
For example, what would happen if every restaurant in NYC cleaned up their kitchens and used such safe practices that they all met the requirements to get an "A" grade on their Dept. of Health inspection? In that case, I would say that they are all safe to eat at, and they all deserve "A" ratings. However according to you, this would mean the scale was broken and we should grade them on a normal curve, giving B's and C's and even F's to perfectly-safe-but-not-quite-as-clean restaurants. This just doesn't make sense - if you meet the requirements for an "A", it's not fair to be put out of business just because someone else spent more time polishing their restaurant floor.
I think school grades should be treated similarly - as an "absolute" measure of whether or not the student learned the curriculum, not a relative measure of how well they performed against their peers. No one should be punished for the "bad luck" of being placed in a class full of wicked-smart students. Now I can't say whether or not the A's given out by Harvard were in fact fair, but the fact that the median grade was an A is not, in and of itself, enough evidence to say that the grading system is broken.
Example: http://www.iihs.org/iihs/ratings/vehicle/v/toyota/4runner
The problem I have is differing GPA distributions between majors. History majors have a much easier time getting in the 3.6-4.0 range than CS majors, because history teachers will rarely give anything less than a B. So when a CS major applies to jobs that other majors are also applying to (say, finance), he can look bad in comparison.