groovin, there are two significant problem with C/R systems. First off, most spam now has forged From addresses, often taken from the same spam list as the To addresses. So if a spammer has your address to send you spam, somebody else is getting spam with your address as the claimed sender - and if they use C/R, you now get spammed with the challenge messages. So in fact now you just doubled the volume of junk email. Second, a lot of users are confused or annoyed by the challenges and will not respond to them. You can miss out on legitimate communication because of them.
That all said, I think C/R has some value and actually do use it myself. The trick is that I use SpamAssassin with its Bayesian learner as a classifier, and identified "spam" gets fed to TMDA for a challenge. This means that most legitimate folks who email me get right through, and mail that gets identified as spam gets a chance at coming back to life instead of being dropped in the void or requiring my time to sort through it. I will not claim that this scheme is perfect - for example it has the forged sender problem. However, so far, this scheme has served me very well, much more so than schemes that I have used before. I get about 300-400 spam messages a day to my main email account, and maybe 5 get through, and to my knowledge no human sent email has ever gotten lost (owing to the C/R providing a way for a human to rescue a false positive).
There are a lot of huge differences of opinion in anti-spam circles because all anti-spam schemes are trade-offs. Therefore, you should never take anyone's opinion - mine included - on an anti-spam solution as being the one and final truth.