That sucks.

When to downsample?

I've been getting good results (mastering) with software: I use T-Racks (tube emulation-eq/compression/multiband limiting/soft clipper) within Steinberg Wavelab 5, as a VST plugin. The last stage is dithering, and I've been using the Waves L2 (limiting/dithering to 16 bit). My question is, when to go from 48K or 96K 24bit files down to 44.1K? Just before dithering? Or should it be done after? I've been downsampling just before dithering, but I'm not clear if that makes the dithering/limiting in the last stage less effective, i.e., lowers fidelity. Or should it be done after the fact, separately, after mastering?

You should dither last.