Since we don't know all that much about human talent - the IQ testers are remarkably reluctant to sort it out into it's component elements (what ever they might be) this probably isn't a well-founded opinion.
Australia from 1942 to 1971, the UK from 1971 to 1993 (in Cambridge from 1982), the Netherlands from (1993 60 2011) and back in Australia since then
Sure they do.
My wife supervised a lot of them - at least 43 - but in psycholinguistics rather than physics. I ran into a few graduate student in chemistry (when I first arrived in the UK as a post-doc) and in physics when I was working on electron microscopes at Cambridge Instruments.
Working out the problem that their Ph.D. project is going to address can be done in cooperation with the supervisor - my wife seemed to do a pretty good job with her students - but supervisors are a very mixed bag.
I'm sure that there are Ph.D. factories which are quite as bad as you claim, but none of the places I knew worked liked that.
Some graduate students subvert their supervisors plans. I certainly did. There was a whole lot of technical stuff in my project that he didn't have a clue about, and while I measured the rate constants that he wanted measured, there was a whole lot of technical stuff and computer processing that he'd didn't know much about.
It certainly doesn't stop good students from finding their own way.
Not in my experience.
There are plenty of bad supervisors. My wife picked up a few of her graduate students from them, over the years. Her first graduate student moved over to her when his - very clever but gay - original supervisor made a pass at him.
There are also a whole lot of good ones. Science funding does get bent by good supervisors to look after good students.