Can maladaptive mutations really be caused by copying DNA that's not used much (as far as we can tell, like the DNA for endogenous retroviruses in our genome)?
The fact that we can tap junk at some future point is probbly just an accidental side-effect... though there is another theory that claims having lots of junk provides some protection against environmentally-induced damage because most of the time it is a junk section that gets damaged. Hows that for the next error protection algorithm: pad the message with mostly zeros so occasional bit corruption doesn't matter. Take that Shannon!
If you want a specific example of this mechanism working: primate 3-color vision. In our two color blue-yellow seeing ancestors the yellow pigment sequence got duplicated, then eventually slightly mutated. That's why the red and green receptors overlap so much yet blue is standing way off by itself. It is high likely this started as a useless duplication and was carried around for a long time before one of the duplicates got mutated.
Obviously, because it isn't junk; it is of value to the organism. Even if it's not of any use right now, even if it's completely biologically inactive at present. Because it is still extremely high entropy information. They're remnants of solutions other living systems once used, at some point, to solve the problem of staying alive.
If I were going to try and exploit genetic mutation to produce novel solutions to biological problems, I would start from an existing genome. In fact, I'd start with as much data, from as many organisms, as I could get my hands on and store. Perhaps we carry junk DNA because mutations in existing coded sequences, even mutated, currently useless ones, are far more likely to be functional, and so potentially a useful adaptation, than literal randomness. It's life's portfolio of solutions, badly photocopied little snippets accumulated over the years, and we all carry it around for future generations that might live in an environment where it's useful.