Ugh that's a small oversight. Fortunately it won't be difficult to go back and rerun all the ones that had a % in the URL. For some reason I was intentionally decoding the URL before encoding, I don't remember why, but probably works to do a single encoding with no pre-decode.

Some of them are not taking.

ID 311991

Source URL (Greek-language characters):
https://archive.is/20130217191756/http://arenalarissa.gr/archives/24525λα-στην-αελ.html

URL Encoded as part of the archive:
https://archive.is/20130217191756/http://arenalarissa.gr/archives/24525%CE%BB%CE%B1-%CF%83%CF%84%CE%B7%CE%BD-%CE%B1%CE%B5%CE%BB.html

URL Encoded a second time for API post:

--post-data='action=modifyurl&urlid=311991&overridearchivevalidation=1&archiveurl=https%3A%2F%2Farchive%2Eis%2F20130217191756%2Fhttp%3A%2F%2Farenalarissa%2Egr%2Farchives%2F24525%25CE%25BB%25CE%25B1%2D%25CF%2583%25CF%2584%25CE%25B7%25CE%25BD%2D%25CE%25B1%25CE%25B5%25CE%25BB%2Ehtml

The API returns "Error" and the Database shows the archive URL containing the original Greek characters. Maybe it doesn't matter.

Archiveis is probably giving it that URL instead. If it loads, I would ignore it.