How much is co-browsing really helping our users?

I've been reading a few things lately about co-browsing and how much it actually does (or doesn't) improve a patron's experience using web-based chat reference services.  Back in November, at the Virtual Reference Desk Conference, I attended a session entitled "User Satisfaction and Best Practices: Correlation in Live Online Reference."  I blogged the session, and the one thing that stuck out to me was that 3 out of 4 times co-browsing was used in a session, the patron replied dissatisfaction with the service.  This was much higher than dissatisfaction ratings for other reference behaviors used (conducting follow-up, naming used sources). 

Then I read the Lethal Librarian's summary and discussion of a session she attended at VRD: "Does co-browsing enhance instruction in virtual reference?".  Different session.  Same results.  From Rikhei/Laura Harris's discussion:

61% of the times the librarians tried to use the co-browsing feature, they had problems with it. I’m frankly amazed at this. How many times do you keep trying something that doesn’t work right?

Personally, I stopped trying to use co-browsing a long time ago.  It's bad customer service to give something to someone that has a good chance of not working.  Period.  Everyone touts co-browsing as the cat's pajamas, and it is kind of cool A) when it actually works, which is rare and B) when the user's question actually warrants it. 

I hear again and again in these two studies and others that oftentimes users get frustrated by librarians trying to teach them instaed of simply pointing them to the resource or answer.  If my user wants just an answer, that's what he or she is going to get.  I always try to explore the teaching angle, but if the person isn't interested or in a hurry, I respect that.

I am not a co-browsing fan.  At all.  If the big-name software companies can get to work on a Mac with a firewall and a dial-up connection running Firefox, then rock on.  I'll be a fan.