diff --git a/python/lsst/analysis/tools/tasks/gatherResourceUsage.py b/python/lsst/analysis/tools/tasks/gatherResourceUsage.py
index faccc1f7d..29087c830 100644
--- a/python/lsst/analysis/tools/tasks/gatherResourceUsage.py
+++ b/python/lsst/analysis/tools/tasks/gatherResourceUsage.py
@@ -617,6 +617,8 @@ def __init__(
                 gather_task_label, gather_dataset_type_name = self._add_gather_task(
                     pipeline_graph, input_metadata_dataset_type
                 )
+                if gather_task_label is None or gather_dataset_type_name is None:
+                    continue
                 metadata_refs[gather_task_label] = refs_for_type
                 consolidate_config.input_names.append(gather_dataset_type_name)
         pipeline_graph.add_task(
@@ -637,18 +639,19 @@ def __init__(
         # to do that again in process_subgraph, even though that's where most
         # QG builders do their queries.
         self.gather_inputs: dict[str, list[DatasetKey]] = {}
+        self.existing_inputs: list[DatasetRef] = []
         for gather_task_label, gather_input_refs in metadata_refs.items():
             gather_inputs_for_task: list[DatasetKey] = []
             for ref in gather_input_refs:
                 dataset_key = DatasetKey(ref.datasetType.name, ref.dataId.required_values)
-                self.existing_datasets.inputs[dataset_key] = ref
+                self.existing_inputs.append(ref)
                 gather_inputs_for_task.append(dataset_key)
             self.gather_inputs[gather_task_label] = gather_inputs_for_task
 
     @classmethod
     def _add_gather_task(
         cls, pipeline_graph: PipelineGraph, input_metadata_dataset_type: DatasetType
-    ) -> tuple[str, str]:
+    ) -> tuple[str | None, str | None]:
         """Add a single configuration of `GatherResourceUsageTask` to a
         pipeline graph.
 
@@ -663,15 +666,19 @@ def _add_gather_task(
 
         Returns
         -------
-        gather_task_label : `str`
-            Label of the new task in the pipeline.
-        gather_dataset_type_name : `str
-            Name of the task's output table dataset type.
+        gather_task_label : `str` or `None`
+            Label of the new task in the pipeline, or `None` if the given
+            dataset type is not a metadata dataset type or is itself a
+            gatherResourceUsage metadata dataset type.
+        gather_dataset_type_name : `str or `None`
+            Name of the task's output table dataset type, or `None` if the
+            given dataset type is not a metadata dataset type or is itself a
+            gatherResourceUsage metadata dataset type.
         """
         if (m := re.fullmatch(r"^(\w+)_metadata$", input_metadata_dataset_type.name)) is None:
-            return
+            return None, None
         elif "gatherResourceUsage" in input_metadata_dataset_type.name:
-            return
+            return None, None
         else:
             input_task_label = m.group(1)
         gather_task_label = f"{input_task_label}_gatherResourceUsage"
@@ -689,6 +696,9 @@ def _add_gather_task(
 
     def process_subgraph(self, subgraph: PipelineGraph) -> QuantumGraphSkeleton:
         skeleton = QuantumGraphSkeleton(subgraph.tasks.keys())
+        for ref in self.existing_inputs:
+            skeleton.add_dataset_node(ref.datasetType.name, ref.dataId)
+            skeleton.set_dataset_ref(ref)
         consolidate_inputs = []
         for task_node in subgraph.tasks.values():
             if task_node.task_class is GatherResourceUsageTask: