1. Correct - assuming there's an associated access-list for the traffic. We may static NAT for all ports but normally restrict the incoming traffic to the ports necessary for the service to operate.
2. Your first citation is the most common method. You might want to have same port incoming forwarded to itself on the real ip address if you have other incoming ports on alternate real ports. for instance, you only have one public IP to work with. Server A gets the traffic sent to its native port (say https on tcp/443) on the public IP. Servers B and C also need to serve up https so you give them 8443 and 9443.
Note you said source port and destination port. That's not the correct way of stating it. You should be thinking in terms of original destination port and translated destination port. Source ports are most often an ephemeral tcp (or udp) port - i.e. something 1025 or higher - and dynamically assigned when a given connection or flow is initiated.
