The author is using string functions for single-byte encodings:
$url = substr($line, $lastDelimiterPos + 1 );
$keywords = substr($line, 0, $lastDelimiterPos);
But in WordPress everything is encoded in UTF-8, a multi-byte encoding. So each time you have a character in your keywords that needs more than one byte, something gets lost.
File a bug report and ask the author to read this article. He will use the mbstring functions then. Probably. 🙂