Sure, a diverging spherical wavefront of the corresponding radius is actually the perfect match for a conical waveguide (and the only one), even better than a flat wavefront is for an OS waveguide. There is no doubt about that. That's the reason I would like to try this out. Now I'm searching for any drawbacks. There must be some as I don't see this being done - actually I have never seen this done like that. Either it is too brilliant or too dumb
I agree with Art that I do see these proposals as being ones that integrate the phase plug into the whole waveguide design. I argued in one of my patents that this should be done and could be easily accommodated with replaceable phase plugs.