who can help designing simulation
You described how the model is built. Could you tell us the main idea of your custom post-processing code from which the leaky constant and pahse constant are extracted? I haven't opened your attachments, because I am not in office and don't have HFSS handy.
From your desccription, I can see that your virtual box intersects with the antenna body. Consequently, portion of the metal structure penetrates into the PML media. This might not be a good idea, because the PML doesn't work well for evanescent waves and it is difficult to judge what is the spectrum of wave along the PML-cut-by-Antenna boudary. If possible, I would make the virtual box big enough so as to enclose the whole structure, including the waveguide feed. If not, I would plot the magnitude of the field along some surface near the problematic boundary. These fields should be much smaller than those on the opposite side of the virtual box. The latter is suppose to be the intended direction of radiation.
For the waveguide feed, did you pad it with a metal on the back (namely the feeding port is short circuit on the source side)? Did you plot field of the feeding mode, look at its phase constant/characteristic impedance to see if they agree with those of a rectangular waveguide? I am wondering if the fact that the waveguide feed is submerged in PML has an effect on the calculation of the feeding mode.