Consider the image below. The solid line is my actual power data from some time late last year. The dotted line is the curve from a critical power (CP) model, fitted to my data. A CP model takes two maximum efforts, and it spits out a prediction of the power you should be able to hold at various lengths - it’s accurate from I think 2 minutes out to 30-45 (or so?).
The two circles are the two points used in fitting the model. The model assumes a curve with a certain slope. If you raise my 5-min effort, the slope changes such that the curve’s asymptote will decrease. That asymptote is what they call critical power. My understanding is that it’s usually pretty close to FTP. They aren’t identical. In fact, in one sense, CP is just a mathematical artifact - it’s the asymptote for the power curve, and I don’t think anybody is going to hold their CP for 6 hours.