I would be careful with the "if they do it it must be sensible" reasoning. There are a lot of reasons that modeling firms using homo-economicus and perfect markets doesn't pan out.
It's not that management is dumb, it's that every layer of management is self-interested. Every hierarchy has to contend with principal agent problems. Organizations can do a better or worse job of it, but the incentives are always there and they lead to predictable inefficiencies which are frequently observed. Everyone knows this, but unless you know how to get someone to manage without self-interest, you aren't in a position to solve the problem. Nobody is. That's why it lingers.
You can play semantic games about whether or not it should be called inefficiency -- I don't really care what you call it -- but the RfR / Dilbert dynamic is very real and very prevalent.