What's going wrong is broadly bug 42424 [1]. I think the best solution is to reimplement dpinst.exe, and to that end I've written a replacement implementation, attached here. It's enough to actually install the drivers so that they'll be used when the device is plugged in, but I can't easily test without the device.
Since you already seem to be working with the Wine tree, would you mind applying this patch, rerunning the installer, and trying the native drivers? I'd be interested to see how well they work.