The main problems I had was where you coughed and mumbled the words, like you said the wrong thing and stopped half way through that word and said the right thing (probably could be called, as Wildman said, hesitations). These kind of reminded me that I was listening to a person speaking if that makes any sense. I didn't find the changes in sound levels that bad, much better then in other induction type things I have tried, but this was a problem for other people so you should fix that.

The first problem can be fixed with a sound dynamics effect, and the second with an EQ effect.
If you want I can do it for you.[/b]
I don't think anything like this should be done, it just needs to be recorded straight and simple so that it is more natural. The only changes should be done by the person listening and that is to adjust the equalizer.