Forecasting Issue

[NWecon]

Using EViews 8 -

Despite having a good overall fit and significant coefficients, many of my models have a large gap between the fitted regression line and the actual value for within the last period of actual data for the dependent variable. As shown below.

When I attempt to forecast out-of-sample this creates a substantial "jump-off" between the last actual value and the forecasts based upon the trend-line relationship. This forces me to make adjustments to the forecast value offline. Is there a way to force the forecast to start from the last actual value rather than continuing from the fitted values generated by the regression?

Any ideas would be greatly appreciated.