The X-12-ARIMA currently does not accept #N/A or empty values in the cells range of the input data. To overcome this limitation, you'd have to call RMNA first to truncate any missing value, then manually find the date of the observation with non-missing value.
Furthermore, I sometime insert #N/A at the end of the series to simplify the model's update as new data become available.
In short, the X-12-ARIMA needs to accept our input data even if it starts (or finished) with #N/A, and the start date must be for the first cell in the data series (regardless if it is #N/A or not). Then NumXL, must figure out the proper starting and ending point of the series, corresponding date, and spare the user this extra steps.
For the forecast, the start of the forecast happens after the last non-missing observation, so this would be dynamics as I insert new values.