Like everything it depends on how it's done. Graded clerkships would be fine if the grading was just: "Top 33%, Middle 33%, Bottom 33%, or Needs Remediation" for each target skill like presentations, notes, knowledge, etc... Instead it's "Did the student meet the following criteria: (Massive list of target behaviors that no attending has the time or energy to read)". Would be even more fine if they then normalized grades according to attending averages and then split grades around 30/40/30.
At my school, subjective grading is horrendous. It's basically 30/68/2 for H/HP/P, and you need to hit >4.5 to get honors. Functionally, that just means residency directors see your score as either "better than average" or "might be the worst student on the rotation, idk." And because you need such a high average, you need an attending who understands the averages are usually that high. So you could get 5 evals and 4 are glowing reviews averaging 4.8, but one attending didn't get the memo, barely remembers you, and gives you straight 3s. Now your average is below 4.5 and you might as well have phoned it in. Oh, and of course they made the shelf honors cutoff the 95th percentile, so you can't make up for even a single poor eval by crushing the shelf, it can only hurt you.
I'm somewhat convinced the shift towards subjectivity coincides with a shift in teaching/academic leadership to far less quantitatively driven people. The liberal arts types tend to flock towards teaching leadership like being a dean or directing a clerkship, and they've brought this anti-stratification attitude with them. At the end of the day, we all get stratified. Personally I'd prefer it be by reliable methods.