The reason you would examine normalized power would be to relate the measured quantity (amplitude or power) to that which is injected at the input plane.
The observation point is a bit of a special case and has limited use in terms of normalization. First the normalization of the observation point is normalizing the amplitude not the power. Second the normalized amplitude of the observation point must be considered carefully if used at all. For example if you have a structure that focuses the signal to a point you will see an amplitude greater than one because while the overall power may be the same the intensity at that point has increased.
In your configuration I would only use observation points to check the time domain signal to verify that the field at the observation region has either sufficiently decayed or has reached steady state. I would use an observation line or XY area located at the same location as the point to do the actual analysis.