Grades don't necessarily exist to maximize some spread and produce a nice normal distribution. They could be constructed that way, but the way they tend to be used is in communicating level of mastery with material, and if a course is well-constructed, the instructor teaches well, and the students are engaged and learning, they should generally mostly get "good grades."
The weird thing is, how do you know that Harvard students aren't basically all understanding the material in their courses well? Wouldn't you expect that?
Think about it another way. Let's say universities tried hard to improve their educational effectiveness over decades, and were successful at it. More and more students were understanding the material in courses better. Why shouldn't that be reflected in improved grades?
At some level as you point out what really matters is whether or not a student basically grasped the material well enough, and the grading system in some ways punishes risk taking by grappling with difficult material.
I was a professor for many years, and there were certain universities that were known to have extremely low GPAs (yes this still happens), so we had to factor that into graduate admissions decisions. It did nothing but hurt applicants. Those students had the benefit of coming from schools that were well-known enough to be on people's radars, and for word-of-mouth to afford admissions committees knowing about grading practices. Students from smaller schools wouldn't be so lucky.
On the other hand, GPA might be more predictive of outcomes than standardized tests (https://news.uchicago.edu/story/test-scores-dont-stack-gpas-...) so this discussion might be moot. It doesn't mean that something couldn't be improved about grading but it's odd to level these kinds of critiques at GPA if it's outperforming the biggest alternative available).