Hi Dan,
I’m an engineer on the cBioPortal team, not an analyst, so I recommend you get advice from an active researcher (there are researchers following this forum and may have some advice). The trade-off between using diploid vs all samples is that, in the former case, you are only considering samples that are not over- or under-expressing from a CNA perspective. Of course, the sample size is probably smaller. There may be other nuances, that, as an engineer, I’m not aware of.
I hope this helps.
Best,
-Benjamin