Evaluating a LangGraph application using LangSmith

Question

I am doing a project where I am evaluating an existing single agent architecture with a new multi agent one. I implemented the multi agent one using LangGraph and decorated the existing one so that I get proper logs in LangSmith. My current problem is that for the multi agent solution, I get the correct traces in LangSmith (attached image), but I cannot access all of the nodes using the Run object I get when evaluating using LangSmith. Namely, I can traverse to the Planner and ExecutionTeam nodes as usual, but no further than that. These nodes have no child runs and no output, even though I can see they do in the LangSmith webapp. Is this a bug or am I doing something wrong?