I suspect that for better or worse, the answer is more prosaic than any of those- the finest minds realize the problem, but no one has any better ideas what to

I suspect that for better or worse, the answer is more prosaic than any of those- the finest minds realize the problem, but no one has any better ideas what to do, so they're continuing on in hopes of getting better results eventually. It's a sort of categorical issue- we've tried inexact, scattershot methods that are just barely controlled enough to be called "medicine" (i.e., cytotoxic chemotherapy.) Now we're trying "targeted" methods which tailor medicines to effect cancer in known ways (i.e., "targeted therapy.") What else is there? Those two categories essentially cover the whole paradigm of "giving medicine to patients." (Note, incidentally, that there is other work going on in the field- early detection/screening methods, for instance, or tumor-subtype classification methods, either of which could make the available treatments much more effective in practice.)