Dear all,
I've checked with CST, which gives the same that the analytical results, i.e. 523Ohm.
CST default impedance definition uses the average of the E/H fields in the port surface. If we force CST to calculate the impedance as in HFSS, i.e. using V²/P integration line (Zpv), then one finds 494Ohm, as in HFSS.
So, it appears that HFSS port impedance (either Zpi or Zpv) does not correspond to the modal impedance of the port (the analytic one, labelled as wave impedance in CST).
This is quite important for hollow waveguide users...