NewsBlur tries hard to filter our dupe stories, but if a story is less than 200 characters it doesn’t de-dupe (provided the guid’s are different) because comics tend to get mashed together if we do.
If the text matches 95%, then we de-dupe. Perhaps you see differences in the stories? If you do but it’s things like image urls being different, that’s a strong signal that I need to change how we calculate the 95%. But I’d need to see two dupe stories that are the same, links/images and all, >200 characters.