Can one really derive classical GR completely from a QFT of a spin-2 field? Or only a linear approximation?