When projections are made for the "total" species the extent of the prediction is truncated.
It looks like NA's in any layer result in NA in the output.
Fix by replacing NA with 0 before adding to total and then once all species are added replacing 0 in total with NA.