Great questions!
Yes, that is around 35 users per AP. It's worth noting that we modeled the design without any consideration for MU-MIMO. The real reason for using PMP450m was for the improved uplink link budget in the DFS bands from the beamforming that the PMP450m can do. In theory a PMP450m should be able to load more users than this and still achieve the desired speeds, but it depends on which SMs in the sector are demanding traffic at any given time.
All users are not guaranteed 100Mbps packages. We needed a way to insure that poor performing SMs would not drag down the sector. We advised limiting the speed plans based on modulation rates. We designed for the following speed plans:
8X => Up to 133Mbps packages
6X => Up to 100Mbps packages
4X => Up to 67Mbps packages
2X => Up to 33Mbps packages
1X => Up to 17Mbps packages
(But in general we discouraged 1X and 2X customers and for most of the other towers in the network retained 5.8 sectors on 20MHz channels to offload poor performing SMs.)
This whole site technically uses 10 x 40MHz today, but using the wider channels leaves unused gaps in each band. (That is to say the 40MHz channels do not fit perfectly in each band.)
It's also worth noting that if someone is a current PMP450i customer and hasn't jumped to PMP450m yet then they can try this technique of using the PMP450m in DFS bands for the greater ranges and purchase a PMP450m limited (which has MU-MIMO turned off but the beamforming is still enabled.) MU-MIMO can be turned on in the future (as needed) with a license key. It also has a 30 day trial as well.
I believe this could be done with ePMP but there are some differences that the operator will have to consider.
1) Max predicted throughput downlink for ePMP(1000/2000) on a 40MHz channel in the 5.4 band is 148Mbps vs PMP450 208Mbps.
2) PMP450m has significantly more uplink beamforming gain than PMP2000 so overall range may be decreased as the DFS bands tend to be uplink limited in link budget.
3) Larger guard bands may be required between sectors that face the same direction. (A tower with 360 degree coverage may not have an issue as channels can be alternated to account for the guard, but an example like this where all sectors face the same direction will be more difficult.)
With that said I think everyone is excited to see the amazing advancements that the ePMP team is doing with ePMP3000. With that platform I believe this type of design will be significantly easier.