ML_norm not working as expected when a scanner with an odd number of crystals per block

[danieldeidda]

a strange pattern is output in efficiencies and geometrical factor even when the model and the measured data are filled with 1s