Most microwave installations that uses coax are split-mount installations, where the actual bit that generates the RF signal is sitting on the mast. The coax carries a lower frequency signal up the mast - it doesn't contribute to the RF noisefloor. Usually the ODU (outdoor unit) sits directly on the antenna. Where the ODU can't be directly on the antenna, waveguide is usually used between ODU and antenna.
Their horn feed statement is also a bit skimpy. Horn feed antennas does have sidelobes, though they're generally pretty small compared to the main lobe. If RF Elements have a revolutionary antenna design that totally eliminates sidelobes, how about showing us a radiation pattern of it? And give some specs - front-to-back ratio and beamwidth would both be interesting to see.
Reducing sidelobes will help a bit to keep the noisefloor lower within a specific band. But as long as you have two operators transmitting in the same band towards the same area, they'll interfere with each other. The only real way of resolving noise floor issues in the bands used by the WISPs is to properly coordinate it. But that means an ICASA that actually have some kind of skill and expertise beyond stamping and collecting fees.
Something else that could help is to mandate the use of very narrow beamwidth antennas for any link that e.g. cross a public road. But that would complicate installations, and many WISP installation teams will not be able to properly pan they're link.
Something else that would really help WISPs is to promote the use of E-band. The last time I looked there was no free bands allocated, though the registered bands were pretty cheap. But the equipment is still way too expensive for WISP use.
