We present a global analysis of experimental factors affecting polarization responses in two-photon inverted microscopy. The role of reflection optics and high numerical aperture focusing is investigated in two-photon fluorescence, which can be extended to other nonlinear processes. We show that both effects strongly distort polarization responses and can lead to misleading extraction of molecular order information from polarimetric measurements. We describe a model accounting for these effects and develop a calibration technique for the determination of polarization parameters in the sample plane using two-photon fluorescence polarimetry in liquids.