James, Let me help out here a bit (maybe this helps, or maybe this is more confusing - LOL)
When a spec sheet says 200 Mb/s, you have to look at the conditions the spec sheet says. Normally, that is at highest modulation (which is saying signal to noise ratio), at the widest bandwidth configurable. When the equipment is installed in the real world, We may be using a narrower channel than our max spec, we may not have all of our SM’s capable of 30 dB SNR. To approximate your net throughput on a sector, you need to calculate the weighted average of the SNR’s that your SM’s are reporting. Then you look to the spec sheet to see what the throughput is at that calculated SNR and at the channel size you are using. Let’s call this value ‘net throughput’.
In an example of no specific equipment, lets suppose our spec’d throughput is 400 Mb/s @ 80 MHz channel. Max modulation level is QAM256, which needs > 28 dB SNR to support that bandwidth. Our spectrum won’t let us use 80 MHz channels - too much interference, so we are using 20 MHz channels instead. That means our new max throughput is 100 Mb/s @ 256QAM: 1/4 capacity for 1/4 bandwidth. We calculate our weighted average of SNR at our known SM locations - our calculation says 18 dB. Our throughput table says that we can support 260 Mb/s @ 80 MHz @ 16QAM, which requires 13 dB SNR, and QAM64 requires 19 dB SNR, so our actual performance will be somewhere in the middle. Back to our 20 MHz channel, and our 260 Mb/s is actually 65 Mb/s @ 20MHz @ QAM16.
In the ePMP1000 spec sheet, it says that expected t/p is more than 150 Mb/s @ MCS15 (which is QAM64 5/6) @ -66 dBm RSSI. The theoretical noise floor on a 40 MHz channel is approx. -92 dBm (thermal noise floor + noise figure of the ePMP1000). -66dBm minus -92dBm = 26 dB SNR (that is signal above the noise). The sheet also says that MCS0, which is BPSK, or 2QAM, needs -90 dBm, or about 3 dB SNR. You can create a graph that has MCS on the vertical axis, and SNR on the horizontal axis, and you can estimate what your net throughput will be from your weighted average SNR calculation. Without going too deep into modulation theory, QAM64 supports about 5 bits/second/Hz throughput (that is where the 200 Mb/s estimate for ePMP1000 comes from: 40 MHz bandwidth x 5 B/s/Hz), and BPSK supports about 1 bit/sec/Hz, which is about 40 Mb/s.
This calculation is the sector aggregate throughput: DL + UL = aggregate. Since the ePMP is time division duplex, we spend some time in DL (AP to SM communications) and some time in UL (SM to AP communications), controlled by GPS synchronization, so the expected throughput in each direction is a function of the DL/UL ratio you have selected.
I hope this helps.