Error found - [Perfect out-of-sample-fit in a model containing a lagged dependent variable?]

Hi there!
Thanks for helping! Patrick was right: I predicted the future by knowing 
the future. The lag-operator is not applied to the model-data's 
dataframe for reasons I'll find out. In effect the model is not Y ~ 
lag(Y) + lag(X1) + lag(X2) but Y ~ Y + lag(X1) + lag(X2)...

Thanks again + good night:
Gero