Limits on differential hysteresis in an enge split-pole spectrograph and on the accuracy of proton beam energy calibrations