diff --git a/backend/Property.py b/backend/Property.py
index cee82836..8f9637a9 100644
--- a/backend/Property.py
+++ b/backend/Property.py
@@ -611,7 +611,10 @@ class Property:
 
             if self.data[description] in self.DATA_ANOMALY_MATCHES:
                 template = cleaned[description][0]
-                fill_dict = dict(zip(template.keys(), [None] * len(template)))
+                # Handling edge case for walls
+                fill_with = False if description == "walls-description" else None
+                fill_dict = dict(zip(template.keys(), [fill_with] * len(template)))
+
                 fill_dict.update(
                     {
                         "original_description": self.data[description],
diff --git a/etl/epc/Dataset.py b/etl/epc/Dataset.py
index 74dcfc56..01391874 100644
--- a/etl/epc/Dataset.py
+++ b/etl/epc/Dataset.py
@@ -498,56 +498,43 @@ class TrainingDataset(BaseDataset):
         Drop properties that have inconsistent data, i.e. changing material types
         """
 
+        starting_and_finishing_null = (
+            expanded_df["original_description"].isin([None, ""]) &
+            expanded_df["original_description_ending"].isin([None, ""])
+        )
+
         if component == "walls":
 
             expanded_df = expanded_df[
-                (expanded_df["is_cavity_wall"] == expanded_df["is_cavity_wall_ending"])
-                & (
-                    expanded_df["is_solid_brick"]
-                    == expanded_df["is_solid_brick_ending"]
-                )
-                & (
-                    expanded_df["is_timber_frame"]
-                    == expanded_df["is_timber_frame_ending"]
-                )
-                & (
-                    expanded_df["is_granite_or_whinstone"]
-                    == expanded_df["is_granite_or_whinstone_ending"]
-                )
-                & (expanded_df["is_cob"] == expanded_df["is_cob_ending"])
-                & (
-                    expanded_df["is_sandstone_or_limestone"]
-                    == expanded_df["is_sandstone_or_limestone_ending"]
+                starting_and_finishing_null | (
+                    (expanded_df["is_cavity_wall"] == expanded_df["is_cavity_wall_ending"])
+                    & (expanded_df["is_solid_brick"] == expanded_df["is_solid_brick_ending"])
+                    & (expanded_df["is_timber_frame"] == expanded_df["is_timber_frame_ending"])
+                    & (expanded_df["is_granite_or_whinstone"] == expanded_df["is_granite_or_whinstone_ending"])
+                    & (expanded_df["is_cob"] == expanded_df["is_cob_ending"])
+                    & (expanded_df["is_sandstone_or_limestone"] == expanded_df["is_sandstone_or_limestone_ending"])
                 )
                 ]
         elif component == "floor":
             expanded_df = expanded_df[
-                (expanded_df["is_suspended"] == expanded_df["is_suspended_ending"])
-                & (expanded_df["is_solid"] == expanded_df["is_solid_ending"])
-                & (
-                    expanded_df["another_property_below"]
-                    == expanded_df["another_property_below_ending"]
-                )
-                & (
-                    expanded_df["is_to_unheated_space"]
-                    == expanded_df["is_to_unheated_space_ending"]
-                )
-                & (
-                    expanded_df["is_to_external_air"]
-                    == expanded_df["is_to_external_air_ending"]
+                starting_and_finishing_null | (
+                    (expanded_df["is_suspended"] == expanded_df["is_suspended_ending"])
+                    & (expanded_df["is_solid"] == expanded_df["is_solid_ending"])
+                    & (expanded_df["another_property_below"] == expanded_df["another_property_below_ending"])
+                    & (expanded_df["is_to_unheated_space"] == expanded_df["is_to_unheated_space_ending"])
+                    & (expanded_df["is_to_external_air"] == expanded_df["is_to_external_air_ending"])
                 )
                 ]
         elif component == "roof":
             expanded_df = expanded_df[
-                (expanded_df["is_pitched"] == expanded_df["is_pitched_ending"])
-                & (expanded_df["is_roof_room"] == expanded_df["is_roof_room_ending"])
-                & (expanded_df["is_loft"] == expanded_df["is_loft_ending"])
-                & (expanded_df["is_flat"] == expanded_df["is_flat_ending"])
-                & (expanded_df["is_thatched"] == expanded_df["is_thatched_ending"])
-                & (expanded_df["is_at_rafters"] == expanded_df["is_at_rafters_ending"])
-                & (
-                    expanded_df["has_dwelling_above"]
-                    == expanded_df["has_dwelling_above_ending"]
+                starting_and_finishing_null | (
+                    (expanded_df["is_pitched"] == expanded_df["is_pitched_ending"])
+                    & (expanded_df["is_roof_room"] == expanded_df["is_roof_room_ending"])
+                    & (expanded_df["is_loft"] == expanded_df["is_loft_ending"])
+                    & (expanded_df["is_flat"] == expanded_df["is_flat_ending"])
+                    & (expanded_df["is_thatched"] == expanded_df["is_thatched_ending"])
+                    & (expanded_df["is_at_rafters"] == expanded_df["is_at_rafters_ending"])
+                    & (expanded_df["has_dwelling_above"] == expanded_df["has_dwelling_above_ending"])
                 )
                 ]
 
@@ -677,7 +664,6 @@ class TrainingDataset(BaseDataset):
         }
 
         for component in components_to_expand:
-            # TODO: change cleaned dataframe to have underscores instead of dashes
             if component == "main-fuel":
                 cleaned_key = "main-fuel"
                 left_on_starting = "main_fuel_starting"
diff --git a/recommendations/WallRecommendations.py b/recommendations/WallRecommendations.py
index 53a74534..c54c00d9 100644
--- a/recommendations/WallRecommendations.py
+++ b/recommendations/WallRecommendations.py
@@ -163,9 +163,8 @@ class WallRecommendations(Definitions):
         if (
             (insulation_thickness in ["average", "above average"])
             or self.property.walls["is_filled_cavity"]
-        ) and (
-            "cavity_extract_and_refill"
-            not in measures
+            or self.property.walls["clean_description"] is None
+        ) and ("cavity_extract_and_refill" not in measures
         ):
             return
 
diff --git a/recommendations/recommendation_utils.py b/recommendations/recommendation_utils.py
index adbeecf5..0794013e 100644
--- a/recommendations/recommendation_utils.py
+++ b/recommendations/recommendation_utils.py
@@ -201,6 +201,11 @@ def get_wall_u_value(
             )
         )
     else:
+
+        # Handle rare edge case
+        if clean_description == "":
+            return 0
+
         mapped_description = epc_wall_description_map[clean_description]
 
         mapped_value = wall_uvalues_df[