I think I follow what you are saying. When you add the entire srvice in the builder you are actaully adding a Dynamic Map service (esri:ArcGISDynamicMapServiceLayer). When you add just the child layer you are adding a Feature Layer (esri:FeatureLayer). These layer types are handled differentl from one another. The Dynamic Service layer is drawn on the server where as the Feature Layer is drawn by the client. the Dynamic Service Layer is much faster on the client side. I have about 100,000 points fairly quickly. With the Feature Layer I limit my services to about 5,000 so the client doesn't hang waiting for the features to draw.
Unfortunately the Dynamic Service Layer is limited in what you are actually able to configure. i belive Pop-Ups are the only thing you can configure seperately for each layer when working with a Dynamic Map Service Layer. I hope this helps...