The idea of a converged mesh is a bit of a misnomer...what you want is a converged result, ie S parameters, Gain, etc. When a result stops changing within a given tolerance from one mesh to another mesh with more density (or higher basis order) it is considered converged. The resulting mesh can be considered a converged mesh though, again, the goal is to converge the result, not the mesh directly. This is especially true since different results can require different meshes to be considered converged...ie S parameters tend to converge more quickly than, say, antenna sidelobe levels...