Marfitalu: Good question, there is no outstanding reason for it, the main reason was simplicity.
I thought about three possibilities :
- The more natural one seems to consider castling as a king move. Then castling is ambiguous can always be replaced by a rook move, and the right to castle would in fact be a disadvantage for the player who can castle. For this reason I don't like this one.
- Consider castling as a two-piece move, hence unambiguous. This is now perfectly sound. But we would also have to state whether it is possible to castle under or through check. As the straightforward set of rules state that check does not exist and that the goal is to take the king, one would have to allow that. Personally I don't like at all the possibility to castle under or through check, like it is the case here in Atomic and Extinction. But this is probably a matter of taste.
- Banning castling is simple, clear-cut, easy to implement and can be phrased in very few words. That is what I like about it :-)

PS Someone composed a retrograde analysis problem of Unambiguous Chess, a variant I invented before Ambiguous Chess, where ambiguous moves are simply illegal (this variant is less playable but fun for problem composition). He asked me whether he could state that castling was an allowed unambiguous move, as his problem needed it, and I told him it was OK with me.