Not only retrovirus and transposon activity, but base substitutions and indels as well. One of the potential functions of "junk" DNA is to act as a sponge to soak up mutations.
I don't see how that's meant to work. The mutation rate is usually given per base pair, and given the way in which copying errors occur, we might expect this choice of units to reflect the nature of that process. The "junk" would only "soak up" such errors if there were a fixed number of them per organism instead.
Am I missing something?