more than 10 years ago some company was doing adaptive beam forming WiFi routers/switches, and what i remember reading about them is that actual direction of the beam for a given connection wasn't necessarily the geographical straight to the other end of the connection, instead it was whatever direction was adaptively found, and that frequently meant very non-straightforward mulitply reflected path. The beauty was that beam power at the "total multi-segmented reflected/bounced" length of the path could still be much higher than onmidirectionally transmitted power even at the lesser distance.