You’re assuming that it’s the same people downloading the free songs from the website as buying the tracks on iTunes. My suspicion is that those are probably different sets of people. The free track-accumulators are active music lovers who spend their time looking around the web for music they like, possibly reading, even writing, music blogs and such. The existence of the free mp3s makes it possible for the music to flow through the online promotional system powered by these people.
These people are also the core drivers of word of mouth, so the artists they like spread beyond the net music-savvy as well. Once it does so, it reaches a broader audience who isn’t as interested in music. This audience doesn’t have the time inclination, or skills to roam the four corners of the blogosphere looking for free mp3s. So they just go straight to the centralized source: iTunes and buy. However, their purchases have no secondary effects. No one asks them for recommendations about music. The don’t post their mp3s on blogs or message board as part of positive reviews. The only promotional effect their purchase has is to increase the artist’s position on the iTunes charts, which is miniscule.
Hence, if you kill the free mp3s, you kill the engine that drives iTunes sales. And iTunes sales are not self-sustaining on their own. The people who buy them don’t generate enough new buzz energy to overcome the natural inertia against the adoption of the new.
I don’t think you need to assert the much less common tip jar behavior in order to explain this phenomenon. I’ve seen the tip jar behavior in action, but I think it’s the exclusive prominence of the more active (and much less numerous) online music digerati. (I’d be surprised if there were more than 400,000 people total who’d ever “tipped” for music that they’d already gotten for free online.)