>If you are already familiar with OSS, but so far never found anything "interesting", the best thing to do IMHO would be starting your own
I couldn't disagree more. If you haven't found a project that is "interesting" or simply useful, then you're not really programming (or perhaps aren't looking in the right place?). There are SO many useful projects, or semi-useful projects (and lots of not very useful ones) that there's got to be at least one (if not dozens) that would be interesting.
Regardless, if one couldn't find an interesting one to contribute to (at least a little bit), starting a new one would be a bad idea because, unless it's really unique or compelling, won't get a lot of attention and you won't get any good feedback on what you're doing (which is critical if you want to learn).
I do agree that the social aspects of contributing to and/or running an OSS project is important. So even if you're not contributing code to a project, if you're filing issues or making comments on other issues, that's a great way to learn.