How to explain this big error of HFSS results

Attached is the HFSS model of antenna isolation. We can caculate the impedance matrix using ideal dipole formula first, and then get the S_ij of each dipole. This results had also confirmed by IE3D.
But the results of HFSS had very big error (more than 40dB!) compared with above results. Anyone can help to explain this? Thanks.