That is completely separate from the fact that genomes are very different from standard computer algorithms, and therefore an algorithmic metaphor is invalid. Your 1k algorithmic string example remains pointless.
Well, mainly I was saying that a communications channel metaphor is completely inappropriate. Bully for you if you're not using one
However, if you were inclined to you could treat the mutated and non-mutated genomes as separate messages, in which case they would inevitably contain slightly different amounts of information. Mutations would not count as noise as, technically, they represent an alteration in the message rather than a corruption of it - the goal is not to perfectly reproduce on the other side what went in, unlike in comms.
If you weren't using Shannon's measure, then please describe exactly what measure you
were using because, forgive me, but you seem to be using the Subjective Measure of Information (SMI) which gives exactly the answer you want in a given circumstance...
I guarantee that if you settle on a solid definition of information then there are trivial examples demonstrating increase in information from evolution.
So, of course, my prediction is that you'll never be pinned down on it.