Okay, I see. The problem is that the regex in the CommentPress parser fails to recognise the ol when there is an attribute - in your case the 'start' attribute. This is one of the weaknesses of the parser and one of main the reasons I want to offer a DOM parser as an alternative in the future.
I have added support for this in my GitHub repo. You can download a zip of the plugin from there and the enhancement will be included in the next release, so you don't need to worry about upgrading when I release the next version.