You are currently managing a real-time data processing pipeline using Google Cloud's Dataflow, and you have configured hopping windows to aggregate the incoming data continuously. However, you have observed a significant issue: some data packets are arriving late, yet they are not being flagged as late data. This discrepancy is leading to inaccurate aggregation results in downstream processes. Your task is to identify a method to ensure that the late-arriving data gets accurately captured and assigned to the correct window. How should you address this problem?