Hi Vladimir, the question sounds a little bit ill-posed.
We are talking about the edge-type S-parameter ports, using the CST nomenclature, right?
Then you can use them either as a source OR as a termination. You are able to define the internal resistive impedance to the value you like. 50 Ohm will be typical for a large amount of experimental setups, but it is not necessarily the correct value for your task.
Only if the discrete port is used as a source (activated in the solver setup) it emits power into the system, which then is fixed by the program to the amount of 1 Watt, INDEPENDENT of the resistive value. The program adjusts the voltage/current itself accordingly.
If the port is not used as an active source, it just will be a discrete resistor which is able to monitor both voltage/current/power (which all are directly correlated by the impedance value and Ohms law). If you choose the resistive value in a way that it matches the line impedance of the environment, e.g. a transmission line geometry, you will be able to construct a (close to) reflection-free wave absorber.
CST uses 1 W as normalized input power for discrete ports (and 1V/m for plane waves). Regardless of the port impedance it's always 1 W (unless you make some manual modification).