You assert that it's a binary choice between obnoxious false positives and life-saving potential, but this is because current smoke detector design is at a local maximum which optimizes for life-saving potential. There's absolutely no reason to think that a fundamental redesign of the smoke detector interface (rather than incremental improvements) can both be less obnoxious and equally life-saving.
As a concrete example, I've been trying to get a friend's parents to install (!) a smoke detector for /years/. They cook incredibly smoky food (lots of yogurt-marinated meats on a dry cast-iron pan) and both they and most of the other members of their immigrant community have removed any smoke detectors near their kitchen because the false positives are so onerous. If it were possible to provide them with a device which they could easily disable during cooking, they would be more protected for the rest of the time.
Additionally, the specific statistics you cite are not designed to determine whether people are disabling their smoke detectors due to false positives, and so have no bearing on that component of the discussion.