Reno,
I may disagree with your analysis.
The original question is not clear; just what does one mean by 'A phase = 2568w'?
If this means that there is a 2568w load from supply leg A to supply leg B, and similarly the B wattage is B to C, and C wattage is C to A, then I'm pretty sure that your analysis is correct.
But if the _total_ load connected to leg A is 2568w, including A to B loads, A to C loads, and A to N loads, then your analysis is not correct, and in fact I don't believe we know enough to answer the question.
Finally, if we are talking all line to neutral loads, with nothing connected leg to leg, then the total power is simply the sum of the power on the three legs.
I completely agree with your basic points: for any 'line to line' loads, you get a certain amount of 'double counting' going on.
-Jon