I suspect you might get a good set of coefficients for your particular crystal with a one-off measurement. If the accuracy of the internal temp sensor is that good with averaging then just heating it up while locking the receiver to a known fixed transmit frequency might suffice (assuming the crystal temperature and radio are well matched, and any absolute temperature offset calibrated).
Mark.