Are you using spectroscopic or single wavelength ellipsometry? I have mainly experience with single wavelength ellipsometry, which in the case of our samples (lipid bilayers on Si as well as other substrates) tended to overestimate the thickness while underestimating the refractive index (since those two parameters have unfortunately quite high degree of interdependence in the fitting model). The mass of the adsorbed layer calculated from product of those two parameters was the most reliable output and agreed very well with expectations (see https://www.researchgate.net/publication/8200985_Surface-dependent_transitions_during_self-assembly_of_phospholipid_membranes_on_mica_silica_and_glass?ev=pubfeed_coauth).

Spectroscopic ellipsometry yielding many more independent parameters gave us values of refractive index and thickness which agreed very well with our expectations. How many layers do you have in your model? Si is under ambient conditions covered by a layer of silicon oxide. So in order to get more accurate parameters of our layer, we had to first determine the thickness and refractive index of the silicon oxide layer to include it as an additional layer to our model.

Article Surface-Dependent Transitions during Self-Assembly of Phosph...