Okay, I see that this is a difference in behavior between vBulletin and XenForo.
In vBulletin, you could enter #'d HTML entities, and if they existed outside the ASCII range, then it would render fine. Inside the ASCII range, and it would be escaped.
However, in XenForo, all #'d HTML entities are escaped before render. As they know they made this change, they designed their importer to handle it fine, by translating non-ASCII entities to their respective code-points beforehand. This change was unexpected for us, and unfortunately in order to make this kind of translation, the VaultWiki importer needs to be refactored yet again.
All imported text needs to be wrapped to account for this difference in behavior.