I suspect that for one the fields were zero at the top and bottom edges of the aperture, and for the other they are not.

The shape is similar to what you would get from a circular waveguide or horn in the fundamental mode, where the fields have finite values at the top and bottom edges but not at the side edges.

@Malcolm White Thank you for your answer. The previous problem may have been caused by the wide antenna beam. Now I have encountered a new problem, that is, the calculated far field is quite different from the theory when the theta is relatively large.

I expect it is the same effect, but a smaller field step at the edge, or perhaps not quite the same discontinuity in field gradient at the edge?

If the gain is about 10 dBi then -30 dB on the polar diagram is -20 dBi, so not a lot of power.

Malcolm White Thank you, the reasons I summarized are as follows: on the one hand, due to the large beam width, the energy at the edge of the sampling plane is still high, and on the other hand, because Ex and Ey were not mapped to Etheta and Ephi.

Well done