Regarding the noise issue, there's nothing wrong with the IFF format itself. Obviously 8-bit sound has a rather limited resolution and you will notice quantisation noise with any 8-bit linear sound format (note that there are 8-bit encodings that are logarithmically delta encoded, which provide better SNR (signal-to-noise ratio) but aren't really supported as general sample formats).
The chances are that many .iff sound files you come across have been captured on an old 8-bit parallel port sound digitizer (or so we used to call them in the good old days). They didn't generally have the best SNR properties to begin with and the noise level will be in the least significant bit(s) of the sample. Add to that the fact that a lot of those sounds were captured at, or resampled to the old pro-tracker C-3 note playback rate (which is about 17kHz IIRC), then you get aliasing noise too. End result are audible hiss, ringing and other artifacts.
If you've used a 16-bit .wav or .aiff sound source, the chances are it's been captured on a reasonable recording device, perhaps even professionally, at a CD quality or better sampling rate. Scaling down to 8-bit adds quantisation noise again, but the original signal noise, which should have been in the least significant bit(s) of the 16-bit data, is now off the small end of the scale. That is to say, none of the noise in the original 16-bit recording, assuming it was recorded properly, can realistically be present in your 8-bit version as it doesn't have the precision to represent it. Only quantisation noise should be present. If you downsample the rate properly (using antialiasing) from 44kHz to say 22kHz - conveniently close to the F-3 note, you shouldn't get too much audible aliasing noise either.
Note that you can use 44kHz samples anyway, if you intend to use mix mode. It's nice if you want to do a direct-to-disk version of a track, since the extra quality is noticable in the end result.
In short, an 8-bit sample downsampled from a 44kHz 16-bit source will always sound cleaner than one sampled on an 8-bit parallel port
