Complex Fenestration Systems (CFS) are widely studied to help redirecting daylight into a room in order to optimise its luminous properties. Daylighting designers and researchers usually employ physical and virtual models to assess their daylighting performance within buildings. However these models do not necessarily provide a sound analysis of real building performance due to a multiplicity of possible sources of errors. Physical and virtual models of a simple office room equipped with a CFS were investigated using comparison with a real building (1 : 1 scale test module), computer simulations based on the Radiance lighting program, as well as monitoring of 1:10 scale models. The main sources of errors were quantified for each CFS on the basis of the observed relative divergences. Significant errors sources, such as surface reflectance, glazing transmittance, CFS modelling techniques and other models construction details were identified: they are expected to be used as guidelines for CFS modelling in the course of daylighting performance assessments.