You might want to reconsider using dither at all when rendering to MP3. To WAV, yes you generally always need to dither, depending on the final stream in Wavelab and target bit depth, but MP3, you really shouldn’t need to.
Also you’re right, Wavelab used to decode MP3 to 16 bit PCM. Now it decodes MP3 to 32 bit float PCM. That’s why you’re seeing a difference.