I'm saying that a smaller amount of data means more compute is required for a join. Sorry if that wasn't clear.