Why would you need to? Those chars should never be generated as part of a recovery code?
That will teach me for trusting my helpdesk when the swear a user is correctly reading something.
Never say never - it's not inconceivable your helpdesk is the first to find a bug, but it's unlikely in this case - those chars should not be created by a v6 client.
Now, a v5 client will create them - but you can't recover a v5 client from EPO....