The question resulted in an interesting discussion, so I guess it must be a good question. But I like neither the question nor the answer (despite having got it right by assuming that the answer would not only be be as stupid as the rules suggested in BoL but would also so interpret the rules as to maximise the stupidness)

And transactional replication certainly doesn't create any extra columns by magic (at least in didn't in SQLS 2000), and neither does snapshot replication; merge replication may do it (I've never used it, so never read up on it).