I don't know why it's bad, but I know the end result is poorly laid out. It's always super obvious I'm reading something passed through calibre conversion, because accent marks will be missing, intra chapter breaks will be missing, punctuation is messed up, etc.
Er, pretty much by definition, when it does a good job you wouldn't know it had been passed through calibre conversion. So how do you count up those "successes" and measure them against the "failures" to determine whether, on a probabilistic kind of basis, it does a good or bad job?
