While I am willing to accept a certain amount of error in count predictions, a five million discrepancy is ridiculous. Either the 'model' is broken, inadequate, or there are other factors such as competence levels or politics affecting the generation of reasonably reliant numbers. I agree that this needs to be looked at and a determination made as to why the forecasts are so consistently significantly off. If the methodology can be fixed, we all will benefit.
Too much water, too little time