Which makes it a curve fitting exercise. They didn't match anything - they changed all the parameters to match the historical dataset, and then declared that no

Which makes it a curve fitting exercise. They didn't match anything - they changed all the parameters to match the historical dataset, and then declared that now the future prediction is correct. Give it 20 more years, they can "recalibrate" again and find another imminent collapse: any model with enough free parameters can match any input you like.