Current interferometers often collect data simultaneously in many spectral channels by using dispersed fringes. Such polychromatic data provide powerful insights in various physical properties, where the observed objects show particular spectral features. Furthermore, one can measure spectral differential visibilities that do not directly depend on any calibration by a reference star. But such observations may be sensitive to instrumental artifacts that must be taken into account in order to fully exploit the polychromatic information of interferometric data. As a specimen, we consider here an observation of P Cygni with the VEGA visible combiner on CHARA interferometer. Indeed, although P Cygni is particularly well modeled by the radiative transfer code CMFGEN, we observe questionable discrepancies between expected and actual interferometric data. The problem is to determine their origin and disentangle possible instrumental effects from the astrophysical information. By using an expanded model fitting, which includes several instrumental features, we show that the differential visibilities are well explained by instrumental effects that could be otherwise attributed to the object. Although this approach leads to more reliable results, it assumes a fit specific to a particular instrument, and makes it more difficult to develop a generic model fitting independent of any instrument.