In the PST you face 3 business cases across 26 questions in 60 minutes (usually when English is not your native language they allow for 15 extra minutes).
If you think about the structure of question for each case, it closely resembles a project: understanding the scope; obtaining results; giving a final recommendation.
According to sources online the passing grade is around ~18 correct answers which roughly equates to 70% passing score.
When you look at number of submissions an office receives per year will be well into the hundreds (maybe even thousands). You obviously need some screening otherwise consultants wouldnt be doing anything other than interviewing people.
So, the screening starts by the CV/CL review, claiming victims ~2/3 or so. Than you move to a math exam which is very cheap for a company (some even have it online like OW), claiming victims ~2/3 of the remaining. So, in just to simple steps we are left with ~10% applicants and we havent even started interviewing anyone. These numbers are a bit exagerated but I would say real numbers would be more on the ~high 20s%.
So coming to your point, PST is so hard to screen out weak performers. You can then rebuttal and say "hey, but maybe some of the people that failed were possible top performers!". True, but when you get hundreds of applicants PER OFFICE PER YEAR you can definitely be willing to lose some.
As a reference point, I'm a senior consultant in a tier 2 consulting firm conducting final rounds in my office and we roughly give 1 offer per 100 candidates applying, MBB would be even lower.
Just my 2cents :)