Can you please elaborate on that?
Contextual ads work well on a site for which the text is predictive of what ads will make money; for instance, in the case of a car forum, you can figure out what kind of cars are being talked about, if they are fixing up drivers, if they are racing, whatever.
Contextual ads work poorly on sites (like Facebook) that aren't really about any topic in particular -- because the text isn't predictive of anything. Also, in some cases, you can predict accurately what ads are likely to make a profit, but the profit isn't much.
Behavioral ads to the rescue -- the ad network sticks a cookie on your and tracks your interests over the long term. Although contextual ads are (overall) better targeted, a network like Adsense that does both can guess which will perform best for a particular impression, raising earnings for publishers and the networks.
Behavioral ads take many forms, but my favorite form right now are ones that are integrated with the shopping process. For instance, if you spend some time on Zappos, you might find ads for Zappos on other sites that feature the sort of products you were looking at. (Right now I'm thinking about what the next Gamecube RPG I'm going to buy next, so it would be logical for AMZN to try to tempt me to close the deal)
Anyway, Google added behavioral targeting and third party ad networks a few months ago -- publishers don't need to do anything special to benefit... we just get bigger checks!
What type of Adsense ads do you use (text, image)?
For one thing, areas that everybody thinks are "monetizable" often have huge amounts of content chasing a small advertising spend, resulting in very low eCPM.
Then there are capricious factors that are very specific to the niche. You do better with a Chevy forum than a Cadillac forum because the real money is in ads for aftermarket parts -- Caddy enthusiasts like OEM parts, so nobody wants to advertise to them. (GM doesn't, since people on Caddy forums are already sold on the brand and know more about the next crop of Caddies coming out than the dealers do.)
Overall, web publishers cheat themselves out of ad revenue by choosing topics the way they do. If you threw a dart at the "encyclopedia of the situation" and picked a random topic you'd get better than average results!
Personally I'm delighted by the increasing use of behavioral targeting technology: I'm finding that behavioral targeting is greatly raising the revenue I get from my worst performing sites.