Without the publishers filtering you can't just pick a book at random and be ensured it probably won't make you want to claw your eyes out (at least publishers would clean up the major spelling mistakes, grammatical errors, and incoherent plots).
So what is left is popularity, and as already established that doesn't stop shitty authors writing books.
Peer recommendations are a real crap-shoot unless you have vetted your peers thoroughly, which takes time.
Whoever can build out an automated system to rank and recommend self-published books will make a fortune.
If you self-publish without doing an ounce of marketing through curation of an audience, then your book falls under the "vanity publishing" sphere. Then why should anyone care if your work is good? You clearly didn't do a good enough job of selling it.
(Note that publishers don't do anything for first time authors in this regard, either.)
Reader reviews are okay for the best sellers if you read a lot of reviews (but I find it tends to heavily skew towards those who loved the book, with a handful of haters), but I find it is pretty much worthless when it comes to average books which only have a handful of reviews.
I'm thinking of a more objective system. It could be based on reader reviews but would need to do something like making the review rank the book on a a set of criteria and then figuring out if the reviewer is trustworthy by cross comparison. eHarmony for books.
1) You are using it in a way that I'm not able to. For you it is a tool that leverages pre-existing relationships, which I'm sure is great. I know about four people that use Goodreads, and their tastes just don't align to me. And unfortunately for me, the stand-alone recommendation service Goodreads offers that people like me try to rely on is just terrible.
2) Even if I had a critical mass of friends making recommendations, they themselves identified the books they should read somewhere. Yes, we can all socially recommend books we like for awhile, but eventually someone has to start separating the wheat from the chaff to start making those recommendations. Books are not movies or songs, and it's much more time-consuming to do this en masse. I don't trust Goodreads to facilitate upstream discovery (that can leveraged downstream via social recommendations) in a publisher-less world primarily populated by self-published books.
2) I'm not sure how much injection of discovery into the network graph in the first place is a concern. There's always people who like trying random new books that they see, or who read something about an author and decide to try them out, or pick up a random book in a bookstore because it looks cool, etc etc etc. I agree that it'll be a much slower process than in the publisher-gated case, but I don't see why you wouldn't see the same sort of ecosystems arise as did for indie music with independent review sites.