The major practical limitation is finding reliable software.  For example, FDTD routines are available for free, but to create your own modeled system, you need to setup the geometries, lightsource profile etc...  thus, if it's really critical, I'd recommend an industrial FDTD solver like the one provided by Lumerical.

If you can model your system as a simple effective medium (ie dielectric constant is homogeneous), you can model it with fresnel's equations.  There are free thin film softwares that do this fairly well.

FDTD is better for standard geometry , but they perform poor in curved geometry , in the latter case go for FEM

Especially, if I want to model something like ATR