TEM cell in CST

Hello,
I try to design a TEM cell in CST but I always get wrong results. I have few questions. Do I have to draw also the air inside the TEM cell or is it enough to draw the metalic eclosure and the septum, as long as I have speciefied the background material as 'normal' (which is something like vacuum)?
The boundary conditions are open in all directions, with little space added in the propagation direction, otherwise I get the warning: "the distance between discrete port and the boundary has to be at least 3 mesh cells". I also have some magnetic filed absorbers inside the cell, so I use subgridding. Is it ok so? I think that I have wrong results (sometimes total reflection) just because of unproper meshing. Are there any "tricks" for a good mesh?
Thanks!
C.

You describe quite compelx model - you include there absorbers. Do you have good parameters for the absorber materials? Can you show your expectations and model?

Thanks for reply! You answered me also some time ago when I tried to simulate the cell in FEKO. I don't have access to it now and I had to quit. As regards CST, what's intriguing me is that the results depend a lot on settings and I don't know which are best. If I use waveguide feeding port I obtain different values than with discrete port feeding. (And yet don't understand the relation between the modes you can define in the waveguideport and the modes that actually appear in my TEM cell)
The absorbers are some ferrites with parameters taken from literature. They were used like that by someone else, in COMSOL, and got results in good agreement with the measured ones. I know the rms Efield values measured in specific points in the center guide, below the septum, and they are my reference. I use subgridding instead of material based refinement, to reduce the mesh cells and, consequently, the simulation time.