Just improved the rule on github: it could miss some articles.
Indeed, I already had the problem before when getting the date from the article links, so the fix was ready: it seems the date in the rss feed is the creation date (when the journalist started writing the article) instead of the publication date. So any article that took more than 2 days to write would be marked as too old and not included even if it was just published.
So now I get all links, and download the article header to get the date before deciding if it is too old or not. A tiny bit slower but we won't miss important articles !
Code still at:
https://github.com/AltGr/Calibre-fre...diapart.recipe