> but there is no evidence that bureaucrats are capable of reliably picking the right candidate drugs

I'm not convinced that the govt bureaucrats would do a worse job than drug co bureaucrats. Current studies are poorly designed, implemented incorrectly, results are cherry-picked and gamed, p-hacking lives, ... This is an industry that can't even seem to accept pre-registration.

Also, as you say, most of these molecules come from publicly funded research. Why would the group funding the first set of research (and producing more molecules than industry) automatically be bad at the second?

In all, I'm not convinced that "Drug discovery is the easy part." If it was so easy, then why don't drug co's do it, and save the licensing fees?