Felix,
From what I understand of VSL's approach to recording their solo instruments, they're not necessarily going for the peak of "realism" with the original recordings at full stereo width, but are rather trying to capture as much stereo information as possible; to have a really robust stereo file, as a sort of starting-point. So the recordings, at full stereo width, will probably always sound overly "wide", and will reveal these sorts of problems more than one might expect. I don't know the details of how the instruments are mic'ed, but I'd imagine Dietz is actually understating it a bit when he says "the front row" - I'd imagine the recorded stereo position would be more like sitting somewhere on the stage! Also, you have to keep in mind that, if you intend to do any imaging of the instruments at all - even for a chamber ensemble - you will still have to narrow the stereo width, as it is certainly impossible for you to be sitting directly in front of _all_ the players in the group. You might be sitting in front of one of them, but the others will invariably be off to the sides.
With the above in mind, however, I do understand your frustration. But I don't think the effect you're experiencing is necessarily related to the stereo width or positioning of any individual sample in the library. I think this is an effect of the Frankenstein-like nature of any sample-playback system. What you are hearing are subtle psychoacoustic effects which are completely natural in _any_ real space, with any real performer, but you're hearing them _without_ the continuity of a live performance. In a live performance, if the musician leans forward slightly, or turns slightly to the side, there is an acoustic continuity to that event, which your ears pick up and adapt to in an entirely natural way. The sound is changing, and the panning effects of the space are reflecting this change, but that change is completely continuous - it never "jumps", because it is part of a physically continuous event. Additionally, everything else you're hearing in that space - the audience, your neighbours, your own breathing - are also continuous, and constantly informing your perception of spatial cues. In a sampled performance, on the other hand, the changes between every sample are abrupt, and that is what is jarring to your ears... (or perhaps I should say, to you mind!)
I have also noticed that this has a great deal to do with the natural decay of the instrument. In the legato instruments, by virtue of the way their recorded, you hear _much_ more of the natural sound of the instrument itself, particularly in terms of decay. Therefor, it's really important to pick your samples and articulations carefully, if you're working on a final output mix. I'm getting into the habit of using performance articulations as much as possible, as these tend to contain the maximum of "real" performance information (transitions, decays), and I also make sure any short note samples (staccato, portato) are allowed to decay fully (that is, either turn up the decay, or lengthen the note in your track). In my experience, this helps to unify the samples of the instrument, somewhat ironically, by allowing the natural panning effects of all the articulations to speak a little more clearly.
Finally, I have to concur with DG on the absolute best solution: narrow the stereo pan, and apply a "small" convolution reverb. This will help localize the sound in your mix, even if you're shooting for an intimate chamber ensemble. In fact, I've just been playing around with the Violin Solo Perf-Legato, hosted in the Vienna Ensemble (which has VERY good control of stereo width and position, btw). I narrowed the width to about 1/4 of full, and added a tiny bit of "small room" reverb and, like DG says, I honestly don't notice the panning effects. And it is certainly not a cavernous, reverberant kind of sound. Adequately intimate, for sure.
Keep in mind, also, that a small chamber group is something like a Turing test for sample libraries, and I don't know of a library that has really passed! You can do **extremely** well with VSL. But a groups of 4 or 5 talented musicians, sat in a room playing your music, will not be replicated by a computer any time soon, IMHO! ;-)
J.
