> For a given amount of data to be transmitted, adjusting bitrate up and freeing the channel quicker is equivalent to using a lower power for more time.
Shanon says that it is an exact equivalence, provided transmit power is >> thermal noise (it is)... If in real world devices that isn't the case, that's a shortcoming of the devices or protocol.
Well, obviously in the real world, devices do not have unbounded transmit rates. But you can get pretty damn close to the transmitter. The Shannon limit isn't the only relevant limit, and is no reason to dismiss out of hand an entire class of optimizations that do help in the real world.
(Also, even if your equipment is capable of hitting the Shannon limit, I'm not convinced that transmitting at the highest power level is going to be optimal for multi-party communications with geographically distributed APs and stations. The information channel between my neighbor's AP and his laptop is neither identical to nor completely separate from the information channel between my AP and my laptop.)
It's not an exact equivalence. It is often possible to reduce transmit power without having to also significantly reduce bitrate. See eg. https://blog.linuxplumbersconf.org/2014/ocw/proposals/2439