It's just the way the popular spreadsheet from velo-fit is coded. CP is held constant once you enter values and curve fit to find CP and AWC. Then that number is multiplied by weight in the predictions section to yield watts at a given weight. So the prediction section basically says that if you drop weight for the same CP, your absolute power at CP(roughly the same as FTP) drops. It's basically the opposite as saying your FTP(in absolute watts) is constant and you'll increase W/Kg if you drop weight.
Anyway, it's just the way they approached this tool. In general all the weight entries should be the same because you wouldn't want to take short and long tests so far apart that your weight changes substantially and you wouldn't want to use old data to predict what your sustainable power is after substantial time(and weight changes) have happened.