I still do a lot of work with older systems like the Kurzweil K2600s where polyphony and RAM memory is at a premium, so I've given this issue a lot of thought. A good stereo recording, even of a solo instrument, contains all sorts of information that our ears have evolved to interpret, and thus is far more realistic. I actually prefer stereo samples that don't have lots of room sound.
For my purposes if I'm creating a brass or woodwind section or emulating a dense orchestration I'm not going to use stereo samples of each instrument. On the other hand if I'm emulating one or two exposed solo instruments I'll use a good stereo sample if available, even if I have to jump through a lot of hoops to make the memory and polyphony available.
When I saw your subject line I thought you might be posting about the need for samples in surround or other multi-channel formats.