Nope, I'm not meaning dithering would narrow the soundstage width, but down sampling by truncating the last least digis do.
Just because e.g. 24 to 16, the last 8bit are lost that containsome parts of the important sound content details those telling our ears and our brains about the important information that how wide are the sounds and the exact postions of the sounds. It's just narrowed a very little bit but not much, even sometimes only when you make an A/B comparative experiment then you would hear out. So just a little wider before the dithering (if necessary) and truncation and yes,
only when going to a
lower bit depth need to widen the soundstage a little bit.
I've tested, just truncate a 24bit audio down to 16bit and the soundstage is really sightly a little bit narrower, but I swear all the sounds, audio sources, track settings, FX effects settings and bus settings are all exactly the same and absolutely nothing different except the truncation (even still without any dithering).
But widening sound image is just a very little bit to just restore the soundstage of the original 24bit to the truncated 16bit output, and really can't over widening, because whe you widen more the sounds, the front-back depth of the soundstage would decrease so be ware to not over used.