one approach, there is a row count data flow task that will add the row count value to a variable for your dataset and you can then perform another conditional split on that value being greater than 0 to hook into the flat file output path.

hth