I have not mentionned pointing accuracy. This problem certainly limits the calibration accuracy in many circumstances. There is no perfect solution to that. Remember to set atmospheric parameters precisely for the refraction correction. Avoid checking the pointing on sources too far away from your object as most pointing errors are systematic effects in azimuth and elevation. The secondary calibration used at Plateau de Bure cancels to first order this problem, since pointing errors should be very similar on the phase calibrator and on the source, but correction for primary beam attenuation can not be applied properly in case of pointing problems.
There are a few circumstances which prevents any good calibration, such as anomalous refraction, high or time varying phase noise in interferometry, or gain losses due to thin partial ice or water coating on the dish surface. The main beam efficiency may drop by a factor 2 in the latter case, and anomalous refraction of more than 10" has been observed also. Unless you can monitor these effects by observing simultaneously a line of known intensity in your source, calibration will be impossible in any of these cases.