I prefer last. If you use the Tube module for distortion, you want to process the distortion, not distort the processor. Or if you use a limiter or compressor, you want to process the compressed signal, not compress the processed signal.
Unless, of course, you want to distort the processor or compress the processed signal
The one exception is if I'm using the CE's transformer emulation and distortion as an effect with bass. Then like any other distortion-based processor, I put it first. But at that point I'm using it more as an effect than for console emulation.
Ultimately...whatever sounds best IS best.