The experimental data collected during the calibration of the polarimeter at JET show clear evidence of non ideal behaviour of the diagnostic optics. The influence of this non ideal behaviour on the FIR polarimetric measurements at JET is investigated, since it results in anomalies which render difficult the interpretation of the detected signals, in particular the Cotton Mouton effect. These anomalies are clearly displayed during the calibration operations in absence of plasma. In fact, when the polarization of the probing beam is rotated, the phase shift between the two detector signals for a particular chord is not constant, as expected, but it changes a lot. After a brief introduction to the implementation of the polarimetry on JET and after a presentation of typical polarimetric signals, the optical characteristics of the recombination plate are analyzed. Their effect is studied using the classical laws of optic. The results show that the recombination plates don't look to be the cause of the detected anomalies. Then, the dielectric waveguides used to transfer the recombined beams from the Torus Hall to the detectors are considered as the possible origin of non ideal behaviour of the diagnostic optics. Assuming that the transmission properties of the optics after the recombination plate are known, a general method to evaluate the signals at the detectors is presented, which includes the effects of the optical components in the dielectric waveguides. As a particular case, the simple rotation of the polarization acquired along the beam line from Torus to the detectors is modelled. The results obtained in this case are compared with experimental calibration data and they indicate that the anomalies present in the calibration data could be at least partly caused by changes of the polarization, due to various reasons such as non ideal components or the reflections on metal mirrors, which may produce rotations of the polarization state of the recombined beams.