It sounds like the interlock protocol only protects against MITMs that try to modify the conversation. It seems likely that most MITMing is for the purpose of merely reading a conversation, not modifying it. The wiki page also describes an attack against the protocol, so it might not be very effective.