JIRA uses the system encoding when parsing CSV files. If the CSV contains any characters that are not valid in this encoding, they are silently accepted and replaced with question marks.

We should rather let the user specify the encoding explicitly in the CSV importer config file.