I kinda feel like these are two different views on the same basic 'mismatch-between-sampling-and-distribution' fallacy, what am I missing here?