The XML data block you found is called a
metadata stream. The data block you called "another set of fields" is called
document information dictionary. Both are documented in PDF spec. (E.g.,
this.)
Metadata stream is richer and more flexible than document information dictionary, and Adobe recommends use of former, but the latter spec. is older and easier to parse, so many PDF handling programs still relys on the document information dictionary. I'm not surprised to know Kindle only support the latter.
IMHO, the problem is not in Kindle but in the program you used to edit the PDF (i.e., PDF-Xchange). It should update both author entries in metadata stream and document information dictionary simultaneously. Adobe Acrobat (an expensive program; not a free-of-charge Adobe Reader) does so.