The routing into streams is done in the “message filter chain” what in your case runs after the processing pipelines.
So if you change the processor configuration order that the processing pipeline runs after the message filter chain you can use the stream rules to route messages into streams and then run the processing pipelines on the data that is in the stream.
currently you would need a processing pipeline that routes the messages into the streams connected to all messages and then a second pipeline that does the extraction based on the data in the streams.