Consider the following text:
At the time of writing, this renders on my system (Windows 7 64-bit, Chrome 29) as:
It looks like 々 (U+3005 IDEOGRAPHIC ITERATION MARK) and 〆 (U+3006 IDEOGRAPHIC CLOSING MARK) aren't being interpreted as kanji (which technically they aren't, I guess, but they behave as one for furigana purposes). If so, fixing this should just be a matter of adding
\u3005\u3006 to all the character classes in
japanese-l-u.js that correspond to parts of a kanji compound (and in particular, 〆 can be compound-initial, while 々 cannot be, as far as I know).