This implies all communication is scaled which is a false assumption. Most interaction is one on one not N to many and there's no reason you need to use the same for both.

Thus conseding interruptions aren't "inherently more efficient", and rather as the author presents different communication methods are more effective under different circumstances.