One more comment: if your Metal1 is a large solid ground plane, I would recommend to modify the stackup and model M1 as a metal boundary, instead of having the full stackup with M1 as a thick metal model.
For RFIC transmission lines above 50-100 GHz, I had seen some strange dips in transmission which we didn't understand for a long time. They disappeared when using M1 as a boundary, and they also disappeared with 3D solvers that model M1 as a solid volume. It seems that this effect is related to Momentum modelling thick metal as hollow tubes, and my port was only touching the top side of that M1 ground hollow tube. Then, at some frequencies two ground current paths (one on the top side of M1 ground, and the other longer path around the edge to the bottom side of M1 ground) created a resonance that isn't real.
This indirectly solves your ground pin shape issue also, because you don't need an explicit ground pin in that case. M1 defined as a boundary will be used by default, as it is the nearest infinite ground.