Exactly. I skim over 90% of my feed marking as read after glancing at it (for high-volume feeds e.g. "planets" the title is enough to know whether I want to read it or not), read 9% of the incoming content directly in the reader, and send the 1% left either to my browser or to a "read later" service for deferred reading later on. My incoming queue gets handled at once, a few times a day.
Shoving all that into a "read later"? Not going to work.
* Sources of headline level knowledge - basically things I note just to keep my brain up on things to search for later should they become relevant. Also useful for the occasional underreported story I'm actually interested in.
* Light reading - there are a ton of comics and blogs that I enjoy for light reading at the bus stop or whatever, enjoyable but really just skimming, or other not-deep forms of reading
* Feeds with occasional good content - basically i don't care about most of these articles, but they provide a gem often enough that I should monitor their content.
And various combinations of these. Product blogs or project blogs are good examples of this. I want to know when the new version is out. I don't care how they solved problem X, but the problem Y writeup may really interest me.
To me, the beauty of Google Reader (and iOS apps like Reeder) is the ability to sift through a lot of incoming information and DEAL with it, e.g. send it to Pocket, Pinboard, Evernote, etc. I'd actually wish Google Reader was even MORE optimized for dealing with large and quick information flows (such as - why can't I feed it Twitter feeds, or any page with/without RSS?). And the reader should do some pre-filtering or guesses on wheter the articles is of interest to me or not.