Talking about the Mesh Fusion examples I had linked in my first post I don’t think that the starting point inside Rhino would differ a lot.
It’s correct that Mesh Fusion uses meshes with proper subD-topology as input geometry. The realtime created mesh however + the baked outcome (with all bevels and rounds applied) is a secondary mesh – it equals the render mesh in Rhino. The topology output by the Mesh Fusion workspace therefor obviously gets critisized by those die-hard mesh modellers who think that only an All Quads mesh is a good mesh… (what’s the whole point, if output sucks…)
The underlying mesh in Mesh Fusion
Similar, newer tools, such as implementations inside Blender and Houdini indeed try sticking closer to the Catmull clark scheme than Mesh Fusion, but they also triangulate borders if need be.
It is generally interesting to watch how the (mesh based) SubD community tries overcoming inherent limitations of the Subdivision Surfaces approach through Live-Boolean Modelling, Face Replacement (> simple faces against pre-made components) and edge roundings, which remain editable.