While you're right about the inconsistency between GR and QM applying at any level, you're wrong about needing GR to talk about the mass of the water molecule.

Even in pure QM, the water molecule will have less inertia than unbonded hydrogen and oxygen atoms. This should in principle be measurable by applying a known force to the water molecule and to the three atoms, and measuring their acceleration. The difference should perfectly match the inertial difference predicted by SR and GR.

GR adds the observation that, if the water molecule has less inertia, it should also bend space-time less, and it is this bending of space time that can't be explained by QM.

Though I should add that I've had a reply to a different comment once that explained that QM is actually compatible with the flat-ish but not perfectly flat space times that GR predicts anywhere not very close to a black hole. They were claiming that in fact modern QFTs can even predict things like the gravitational lensing produced by our sun, and that they only break down when near the event horizon of a black hole.