Forecasting errors: yet more problems for identification ?