I'm not aware of any claim by Meridian that apodizing by itself should be performed as early as possible in the chain. However:
1) Meridian says that it is critical to upsample as early as possible, to reduce jitter in transfers. As apodizing and upsampling occur together, this means apodizing as early as possible.
2) There certainly seem to be different apodizing implementations: 808.2/3, 861v6/G68SL, HD621, and DSP loudspeakers. We don't really know the differences between the filter implementations in each device category, but anecdotal evidence suggests that apodizing in DSP loudspeakers is not as effective vs. apodizing performed earlier in the chain. My gut feeling is that this is due to increased jitter in the 1x sample rate feed into the loudspeakers, rather than a "weaker" apodizing implementation, but I may be wrong.