I'm not Jamie, but as I have some experience with databases, I can tell you that the answer is that it would be quite hard to accomplish what you're seeking. The database can only have a single way of representing "Unknown" because databases work by cross-referencing indexes, and there needs to be a unique identifier representing Unknown for this to be done efficiently.
To accomplish a multi-lingual unknown would mean establishing a second field, so you'd have one field that actually signifies that the last name is unknown and another field for the text that is desired to represent that the name is unknown in the language of the profile manager's choosing. Then you have to write a whole lot of software to synchronize the two values and present a means for people to select how they'd like the last name to read. Then you have to teach everyone what all the different words are that represent Unknown in the various languages that people choose to use.
I support the notion that Wikitree could and should do more to de-Anglicize itself. In this particular case, though, I think you've hit on a really tough spot.
Perhaps a language neutral means of representing it would be better. For example, three underscores ("___") could be the symbol for an unknown last name. To accomplish this, the software would be modified so that everywhere it treats "Unknown" specially, it would instead do that for "___", and all the current values of "Unknown" would have to be updated to "___". This amount of work to accomplish this would still be quite significant, but it would be orders of mangitude less work than trying to make the Unknown flag multilingual.