Bayesian History Start Date

We are using Bayesian Method of Forecasting using Actual Sales and Sales Plan.

Now business has found that for most of these short life-cycle products, the initial 1-2 weeks sales shoot up and then it comes back to normal which is how the sales curve is built. Due to this initial high sales, the Bayesian Forecast gets modified to very high levels which comes back to normal once the initial period has finished.

The requirement is to not let the initial sales affect the forecast. We have a measure 'History Start Date Override' which is used to specify the date from which the history has to be used. Does this measure need to have a value in the elapsed period or can we set up a future date which is 1-2 weeks after the item start date?