Implemented by adding the extra 3-value syntax as its own case and only running it when parsing background-position. I'm sure it could be implemented in a smarter way but this is still a bunch less code than before. :^)