Google Ngram sorting?

From what I understand, each file in Google's Ngram dataset contains a list of ngrams, sorted alphabetically, then numerically by year. However, assuming the data is UTF8 (which file says is correct), и is 1080, where I is 73, so I don't understand why использовал_NUM comes before I'Academie_PRON. The relevant lines from the file (starting from line #131356):

использовал_NUM 2005    4       1
I'Academie_PRON 1813    1       1

Here's my ngram-sort-test.js with the broken comparison function highlighted. To run, download this file from Google and un-gzip it in the same directory as ngram-sort-test.js.

This is not really an answer, but my workaround was to manually sort the file using LC_ALL=C sort <googlebooks-eng-all-1gram-20120701-i >googlebooks-eng-all-1gram-20120701-i.sorted.