The fact that there is good agreement between the model and real world data extending 40 years past its creation should, at least, raise your prior probability

The fact that there is good agreement between the model and real world data extending 40 years past its creation should, at least, raise your prior probability that the model is correct enough to be useful. If not, you're just being irrational.