Just to second what Glassman said, interger upsampling (44.1 to 88.2, 176.4, etc...) is considered better than non-interger upsampling. (96, 192, etc...)
All external variables aside, non-interger upsampling requires some pretty exotic algorithms to produce a perfect conversion. It's possible that this coloration could be "pleasant" to some folks considering their listening preferences. Considering MP3s, coloration can without a doubt make them sound better. Interger upsampling, (as opposed to non-interger) would be less forgiving with lower quality sources.