I believe we both agree. However, when evaluating the costs of the mediating entities such as Elsevier the assumption is that you will be starting from scratch if those entities were to disappear tomorrow, and you will not simply be usurping their brand names/prestige/authority etc. The publishers may be coasting today on their current names - but note the comment in an adjacent thread [1] which talks about new publications and their general challenges - there was a real effort at some point to build up a name they can coast on, and that cost was probably not trivial.
In fact, even though I started the Elsevier bashing in this thread :-) - until we find out what the replacement system looks like I think we may not even completely see all the costs involved. I don't think the existence of the internet is suddenly going to turn the research publication process into a very resource efficient system.
[1] https://news.ycombinator.com/item?id=12436924