If the circle is inside the rectangle, then there is less material, so a physics argument is that it should take less force to deform the circle than the rectangle.

If you get two successful meshes on the same geometry, and one has an edge length of 1.5 mm and the other has an edge length of 1.0 mm, if the deformation remains at 0.35, then that is evidence of a mesh-independent result.  If you repeat for the other geometry and that value doesn't change for two successful meshes, then you can conclude that there is a difference between the geometries.

Ideally, you would also create a mesh with an element edge length of 0.75 mm and a coarse mesh of element edge length of 2.25 mm. Then you would have four points to plot the result vs. edge length and see if the value has converged on the exact value. See this discussion.