The 'waveform' files (.S1M) contains 'Yamaha waveforms', which basically maps key- and velocity ranges to waveforms, much like an instrument in Awave Studio, except there's no associated 'articulation data' (envelopes, lfo's et c). When opening one of these files, each waveform therein will show up as an instrument. When saving, you can save an entire collection, or a single instrument, layer or region and 'Yamaha waveforms' will be created from these.
The 'voice' files (.S1V) contains one or more 'Yamaha voices' which basically correspond to the instruments in Awave Studio. A voice contains from 1 up to 4 "elements" (each one corresponding to a layer in Awave Studio). Each element has a set of articulation data and maps to a Yamaha waveform.
The 'all' files (.S1A) can contain many different types of data; voices, waveforms, performances, sequences et c - but only the waveforms and voices are read.