In my reading of the problem, I concluded that the creation time was also the likely most-recent modification time. Especially in the subsequent clarification of the problem statement, it seemed clear to me (although not necessarily true) that the issue was to avoid analyzing files from previous runs of the procedure. In such a case both the mod time and the inode-change time of old files are likely to have been unchanged during the most recent program run.

I'm all for clarifying that ctime is not the "creation time" but the initial statement here, "That might be impossible to do on a Unix system" really caught me by surprise. Had it not come from a credible source, I would likely have ignored the remainder of the message.