Thanks for confirming. The width on Auto and WPA2+AES is perfectly fine. Auto does 20MHz on 2.4 and 40MHz on 5GHz, which is what you want. Fat 2.4 channels are a bad idea and most clients will even advertise they are fat-intolerant anyway.
So, on 2.4, 40mbit or so is generally the peak I see from 1 stream ("150 mbit") devices, and on 5GHz, I can see 80mbit thanks to 40MHz channels. But that 80Mbit is on a good day with the right testing app and provided the wifi is not dumbly implemented on the device itself (you wouldn't believe how many mobile devices have bandwidth and latency bottlenecks between the wifi stack and the rest of the system)
Have you ever seen better than 40-50mbit from these devices with other AP's, including consumer ones? Do you get substantially worse than 40-50mbit from more than a couple feet away from the AP?
I'd also try to rule out poor client roaming choices. Set up a test environment with just ONE AP, one band at a time, and stand close to the AP and repeat your bandwidth test. If that performs better then it's likely your clients are associating with a far away access point, and there are some ways to work around that.
Honestly, for the devices involved, I think you are seeing fairly acceptable speeds from them. If you want something faster, I would really demand that your clients be MIMO, at least 2x2 (gets you 80-100mbit real throughput) or 3x3 (gets you 150-175mbit real throughput). Or 802.11ac -- if you upgraded to R500/R600/R700, you could see 20-30% higher speeds from the efficient 256QAM modulation scheme, and generally speaking 802.11ac wifi devices are architected to support the faster speeds with the appropriate internal bus architecture that allows those speeds.
P.S. On a related note, if the bulk of your devices you want to support are 2x2 (300mbit) or 1x1 (150mbit), you might want to consider Ruckus's 2x2 stuff (like the R500 or R300 or 7372) since you really aren't getting too much advantage out of the 7982/R700 unless you have a ridiculously dense or polluted environment.