In calorimetric neutrino mass experiments, where the shape of a beta decay spectrum has to be precisely measured, the understanding of the detector response function is a fundamental issue. In the MIBETA neutrino mass experiment, the X-ray lines measured with external sources did not have Gaussian shapes, but exhibited a pronounced shoulder towards lower energies. If this shoulder were a general feature of the detector response function, it would distort the beta decay spectrum and thus mimic a non-zero neutrino mass. An investigation was performed to understand the origin of the shoulder and its potential influence on the beta spectrum. First, the peaks were fitted with an analytic function in order to determine quantitatively the amount of events contributing to the shoulder, also depending on the energy of the calibration X-rays. In a second step, Montecarlo simulations were performed to reproduce the experimental spectrum and to understand the origin of its shape. We conclude that at least part of the observed shoulder can be attributed to a surface effect.