Even in the simplistic way you are looking at it, a mutation by repetition adds information.
No, it adds noise. If adding noise added information, then the more static you add to a message the more information it would have. What you don't seem to understand is that the amount of information from a source is determined by the receiver, not the sender. The question you need to ask is what processes those repeats? Was it modified to understand the static?
It literally adds a substrate for the environment to have "it's" say about. It could possibly be just noise for a while in the gene pool. However, it could alone or in combination with other mutations give reproductive advantage.
If you understand what I just said above, you'll understand why all improvements, bar none, cannot contain too much noise. Which is why all improvements we see so far are subsets of the information that was already there.
It works fine in conjunction with epigenetics.
Ah, yes, the common descentists new black box. This is what you are saying: "since DNA has not been able to make a novel feature, we'll start claiming it's epigenetics that makes magical improvements since it is not well understood how epigenetics works yet."
Epigenetics will be understood some day. It won't save you.
I do not care what you said. I care about only what the experiment means in context.
What the experiment means is exactly what I said, even in context. Both the Harvard experiment and the Lenski experiment show, as in every other similar experiment, that the information in the resulting organisms are subsets of the information their parents had.
If you think otherwise, you'll have to show the context.