Having the pagemap.xml stored inside the mobi and not in a separate file is new (and interesting). I am not sure if it is something that needs to be stripped or not. The issue is the sanity check for SRCS. Perhaps that should allow SRCS or PAGE or maybe we need to be able to extract the PAGE information similar to how we extract the SRCS.
Would you please post a zip archive of a sample epub ebook that uses a pagemap.xml file so that we can run it through the latest kindlegen to see exactly what is being stored in the PAGE section of the mobi and if it is used or referenced anyplace else in the header.