looks like yet another issue. We had this in the past (not exactly the same case but we were overcharged by success evaluation of failed calls) we raised with support and it was fixed in the next release