ä and ß fall within the ASCII range. It would be difficult for them to be misidentified. The characters in your OP are good characters for testing; when testing special character support, I generally use Cyrillic. By the way, these characters pass unimpeded when used directly:
https://www.vaultwiki.org/xf-wiki/in...ράμνιον
If the original content was saved with mixed character sets, as was easy to accomplish in VaultWiki 3, then there is almost no way to guarantee that all characters will be safe.
The importer for VaultWiki 3 assumes that the database connection character set is correct (for a default source vBulletin, blank is correct). When determining the source character set for an article:
1. It compiles a list of all forum languages and their defined character sets and orders them from least-inclusive to most-inclusive
2. It loops through this list and checks for errors in the source text. If there are any errors, it tries the next language.
3. If no languages were successful, the most inclusive character set UTF-8 is assumed.
This means that in order for an article to be imported with any ???, the source article probably contains a character that is invalid for the character set the rest of the article is written in. Such an article has multiple character sets (was probably edited by users who had different forum languages active) and it is not possible to import correctly unless the source article is fixed.
If VaultWiki 3's Multi-Language Support was not active for the entire life of your wiki but your forum contained multiple language packs, the likelihood that such articles exist is high.
But even with Multi-Language Support, any article that was moved from one language to another at any time, or any character that was pasted into an article but that did not exist in the article's current character set, might make the source character set indeterminable and default to UTF-8 (which is also wrong).
However, there is no way for me to know whether any of this even applies to you or where the problem is in your case for this article without knowing which article it is in your source forum and without access to the importer. Even if you were to paste the source text here and it looks the same as on your source forum, the data will not be encoded in the same way here as your importer is receiving it.