Would this be allowed given that I will fit a model based on a model built from future values?

@GraemeWalsh thanks. Lets call the time series in question Y (unemployment) and the other time series as X. All the predicted values have a VAR(3) except the last one (the isolated one in 2011) with a VAR(4). The first 3 predicted points come from a model with difference value of 1 for the X time series. The 4th-7th predicted values come from a model with difference of 2 for the Y time series and 1 for the X time series. The 8th-14th predicted value come from a model with difference of 2 for X and Y. No seasonality detected. Should I keep the starting model for all the predictions?

@GraemeWalsh thanks. The first 3 predicted points come from a model with a

Ok , I edited my question.

You are totally right. I forgot to remove those ...how can I edit the question?