Well, for a few of the un-imported pages at least it wasn't a VaultWiki error. Not sure if you were importing from more than 1 database, but those articles had the same ID number of some of your images that were already imported, but didn't exist in the MediaWiki database.
Anyway, there were some other errors which could have contributed, mainly that the output character set wasn't being set properly for non-default languages (like if you upgraded your board from vB 3 you probably still use ISO-8859-1 rather than the expected UTF-8).
I'm marking this as fixed, because once I told the importer to ignore duplicate IDs it imported the missing pages.