My understanding is that eigenfactor rates journals, not individual papers, so if somehow you get low-quality (whatever you want that to mean) papers into nature it has no independent way to realize that your specific paper is low quality. Also eigenfactor is biased towards favoring larger journals, which is not obviously a good thing. It would honestly be really cool if someone did page rank for individual papers. It seems like a much saner metric than anything that is currently used.