diff --git a/pipelines/HSC/DRP-RC2_subset.yaml b/pipelines/HSC/DRP-RC2_subset.yaml
index 5421e507..aa60ae6b 100644
--- a/pipelines/HSC/DRP-RC2_subset.yaml
+++ b/pipelines/HSC/DRP-RC2_subset.yaml
@@ -400,10 +400,11 @@ subsets:
       - skyObjectStd
     description: >
       Tasks that can be run together, but only after the 'step1', 'step2',
-      'step3', and 'step4' subsets
+      'step3', and 'step4' subsets.
       This step includes patch-level aggregation Tasks. These should be
-      run with explicit 'tract' constraints in the data query, otherwise
-      quanta will be created for jobs with only partial visit coverage.
+      run with explicit 'tract' and 'band' constraints in the data query, or
+      quanta will be created for jobs with only partial visit coverage and
+      for bands with no columns in the object table.
   nightlyStep8:
     subset:
       - validateObjectTableCore
diff --git a/pipelines/_ingredients/HSC/DRP.yaml b/pipelines/_ingredients/HSC/DRP.yaml
index 426a79c4..ea1bdc20 100644
--- a/pipelines/_ingredients/HSC/DRP.yaml
+++ b/pipelines/_ingredients/HSC/DRP.yaml
@@ -72,3 +72,8 @@ subsets:
     description: |
       Set of tract-level faro metrics to be run on coadd products. These use Object
       Tables which are available after consolidateObjectTable (step3) is run.
+
+      These must have a band constraint in their data query to avoid generating
+      quanta for bands that do not have columns in the object table, since
+      using an object table dataset as input does not provide information to
+      the middleware about which tasks went into making the object table.
diff --git a/pipelines/_ingredients/LSSTCam-imSim/DRP.yaml b/pipelines/_ingredients/LSSTCam-imSim/DRP.yaml
index c69d13d9..f869816c 100644
--- a/pipelines/_ingredients/LSSTCam-imSim/DRP.yaml
+++ b/pipelines/_ingredients/LSSTCam-imSim/DRP.yaml
@@ -300,3 +300,8 @@ subsets:
     description: |
       Set of tract-level faro metrics to be run on coadd products. These use Object
       Tables which are available after consolidateObjectTable (step3) is run.
+
+      These must have a band constraint in their data query to avoid generating
+      quanta for bands that do not have columns in the object table, since
+      using an object table dataset as input does not provide information to
+      the middleware about which tasks went into making the object table.