I've processed 17 runs of calibOps configuration by calibGenCAL and have made a fit of a proton peak position for each crystal.

Following two plots show the proton signal for one crystal on orbit


 

and the muon signal in the same crystal on the ground.



 

The correlation of  MPVs  on the ground vs orbit is shown on the next plot for all crystals 


 

The histogram of the MPV ratio orbit/ground on the next plot. 


 

So proton signal on orbit is in average 5% smaller than muon signal on the ground. The simulation of proton signal taking into account the real spectrum, angular distribution both(different from ground muons) is needed to extract correct crystal calibration coefficient (so far I just used 11.2 MeV - the same as for muons).