Yeah, we're working on identifying whether something is in quoted or non-quoted text. That one is quite tricky to get 100% right, so we err on the side of matching more messages.