Regarding the simulation geometry and setup, The setings for the two cases are similar.
The solving space Box1 size: Position: (-500, -500, 0); (Xsize, Ysize, Zsize) = (1000, 1000, 600). It is larger than the suggested size: Lambda/4.
Solution types: HFSS-Network Analysis-Terminal (ANSYS 2022 R2)
Analysis Setup: Freq = 4 GHz, Maximum number of passes: 20, Maximum Delta Energy: 0.1. Others: Default. Sweep: discrete-single point at 200 MHz.
It is also noticed that the sizes of the solving space can affect the result of case 1, but cannot affect the result of case 2. The result of case 2 is constant and is always equal to the one obtained from the theoritical calculations. Therefore, I guess the difference is due to the pec GND of case 1. But I don't know the reason.
I hope the information is enough and I look forward to your feedback/suggestions.
Thanks and best regards,