Sorry for the delay in answering. You raise some interesting issues, not all of which are answerable given the state of the art, as I understand it at least.
On the subject of noise, there are very strong theoretical reasons (e.g. Shannon's theorem) to think that noise in a channel in the same mode as the information modulation is an absolute limit to the rate of data transmission in that channel using that mode of modulation
The bolded caveats are important. Noise present only in one channel doesn't affect information transmission in a different channel (e.g. one channel using radio, a second using optical). Approximately white noise in the amplitude domain has limited impact on frequency domain modulation (to a first loose approximation), which is why phase locked loops manage the apparent magic of being able to track FM signals below the noise floor.
Little is known about sources of noise in this new OAM channel. Certainly nature must produce *some* noise there (due to various fundamental principles), but it is likely IMHO to decrease with increasing OAM energy (see the history of physics topic "ultraviolet catastrophe" to see why). But we have much to learn; we shall see.
As to single photon RF systems and exotic technology, well, we're certainly talking about a very exotic technology here, so I wouldn't worry about exotic-ness. :-)
Finally, as to photon counting versus group ensemble production/detection in other regards, that's something interesting to ponder. I have a dozen half-baked thoughts on the subject, but mostly they are about various kinds of tradeoffs and circumstances, rather than about a rule of thumb concerning maximizing data rate. Instead of trying to to verbalize all that, I'll just say that I don't know.