...
40 deg < theta < 50 deg
55 deg < theta < 65 deg
The visible deviations of fit from the spectrum is explained by low energy tails increasing with theta angle, as confirmed by simulation: 
Using peak position for theta < 10 deg as normalization I found the deviation for linearity as a function of crystal energy (not corrected for angle):
...
I'm developping alternative method, which could be applied to all peaks, including p, He, and Li.
Method is based on energy split between adjacent crystals.
The high energy tail of GCR peak is explained by GCR energy spectrum and rather well reproduced by simulation: 
