I've done this on a project I worked on recently, so I'll take a moment to explain why I think this is a valid, and safer strategy than the alternative.

In said project we had a number of modifications to libraries to support some additional desired functionality. The developers at the time simply forked the library and wrote patches on top of the mainline branch. We found this significantly impeded our ability to upgrade, or take advantage of changes, down the line, as well as to track upstream's HEAD. When one forks, branches, and then patches on the branch, these problems are easily skirted. In addition, it's easy to submit your patches for integration into upstream.

In the end the gain is flexibility. One cannot discount the value of flexibility.