When galaxies show gravitational lensing features, we can use those jointly with stellar dynamics to learn about their mass profiles. When doing that, uncertainties on the orbital structure (mass-anisotropy degeneracy) and deprojected luminous profiles play a non-negligible role. I will show how these can be tightly constrained or even erased using two physically motivated models. The first one relies on suitable integrations of the Jeans equations and produces mass-constraints that depend explicitly just on the surface-brightness and LOS velocities, allowing for a limit where the results are completely independent of orbital structure. I will provide the basic, ready-to-use steps of the procedure and examples on galaxies in the literature (based on work by AA, MWA \& NWE, submitted). The second method relies on a few-parameter class of physically sound distribution functions for both the luminous and dark components, which have already been studied in the literature but never applied to gravitational lensing by galaxies (based on work by AA \& MWA, in preparation).