Well, I remix music in the audio portal so I know quite a bit about audio compression.
Firstly, the conversion of an MP3 to a WAV before putting it in flash is real handy if you want a seamless loop of a song. Flash tends to have a small delay when replaying loops that were in MP3 format. I really don't know why.
Even with my super clear songs, I can render as low as 64 kbps without losing quality. As well, for voices, think of it like a phone line. If you were to call someone, that is usually running at ~16 kbps. A voice clip, at max, should be recorded at 32 kbps. So unless you're somehow singing a song or something (have yet to find a submission that does), 32 kbps is more than enough.
And as well, for people wondering about about Audio Optimization, there has always been a link on the AP main page. But I'll sum it up for you here.
The best way to use audio in a submission is by using the stream feature. In simple terms, this plays the audio until it hits a frame that doesn't have that audio clip set up. Since you cannot change the frames and such in the swf itself, flash automatically cuts down the audio to fit that region, saving you time and size.
As well, this allows you to pause the movie and have the audio pause as well. This is due to a syncronization process built into the swf.
As well, when exporting your movie, under the part about adding a password to the swf and stuff, click the "Overwrite audio settings" or similar button. Then, set a bittrate to your need. Personally, my opinion is that unless you need a 3D sound environment, you might as well go with mono. Anyways, render the movie and test it. If you feel you can up the bitrate a little, just do it and render it again.
That's the just of it.