I don't mean to stick my head into this suddenly, but I wouldn't attempt mastering of an mp3 sound format. Even an mp3 at 320 constant bitrate on the 44.1KHz fq. level . . . although may be useable to practice frequency filtration for overall balancing of the final sound, the bottom line is that you could never take it to the bank for just one reason only:
mp3 is lossy (at any configuration)

Take a wav and convert it to mp3's finest. (using your best Fraunhoffer or Lame or whatever brand codec)
You just created a detailed log of frequency merges which is what makes it so it can be compressed. Any frequency source that matches the frequncy of another source in the wav file are combined and details of it logged into that block of frequency range and then compressed. Some Mp3's are amazingly close out of the speaker, but they are far from the same on the wav/mp3 editor. Take the mp3 and re-convert it back to wav now and compare the statistics. Better yet try cross channel fading on the once an mp3 wav. You cannot do it to the wav that has been converted and re-converted back. You can, but it is horrible sounding because the frequencies were merged during encoding. This is how mp3's digitally represent sound. Technically speaking mp3 is a SOUND format and not a form of preservable audio media. Rather It's just a compressed digital representation of combined frequencies. The theory began long ago when searching for ways to make FM stations more powerful and longer range. The same pricipal goes for the dynamics of the media. try putting some quality compression or other dynamics process on an mp3 same thing.

I have some wavs you can mess with if you want.
MORPH!