The mutations are like little nudges to throw you off the local maximum.
And there's junk DNA where mutations can accumulate over time without being subject to selection before getting enabled at random to see if they give you an advantage.
I think both the amount of junk DNA and the mutation rate are themselves subject to evolution for the best trade-off.
Note that most "junk DNA" is just DNA with a purpose we don't understand (originally, all DNA that didn't code for proteins). Some of it is true junk, of course.