I think I know the answer. In the database where it's working, the language for the index is not Chinese, where as in the one where you don't get a hit, the the language is Chinese.
I ran this query:
SELECT * FROM sys.fulltext_system_stopwords WHERE stopword = N'人'
and it returns to rows. I don't know LCIDs by hear, but I don't think 1028 is Italian. :-)
A stopword is one that is not included in the index. For English this are small and common words like "to", "from", "and" etc for which it would be useless to search because there are so many of them.
Whether 人 are of the same ilk, I don't know, since I don't know Chinese.