diff --git a/polars_vec_ops/frame.py b/polars_vec_ops/frame.py
index ac4b59a..6be2862 100644
--- a/polars_vec_ops/frame.py
+++ b/polars_vec_ops/frame.py
@@ -270,10 +270,8 @@ def _is_sorted(val: Any) -> bool:
             kwargs={"relative": relative},
         )
 
-        # Post-process: convert to counts or cast to correct output dtype
-        if as_counts:
-            cross_clip_expr = cross_clip_expr.list.len().cast(pl.UInt32)
-        elif return_dtype != pl.List(pl.Float64):
+        # Post-process: cast to correct output dtype
+        if not as_counts and return_dtype != pl.List(pl.Float64):
             cross_clip_expr = cross_clip_expr.cast(return_dtype)
 
         # Apply cross_clip to extract only _TEMP_VAL column (avoid gathering large lists)
@@ -281,6 +279,22 @@ def _is_sorted(val: Any) -> bool:
             cross_clip_expr.alias(val_col_name)
         )[val_col_name]
 
+        # Convert list results to counts after clipping, preserving nulls.
+        # This is done after the clip rather than inside the expression because
+        # list.len() returns 0 for null lists in older polars versions.
+        if as_counts:
+            clipped_series = (
+                clipped_series
+                .to_frame()
+                .select(
+                    pl.when(pl.col(val_col_name).is_null())
+                    .then(pl.lit(None).cast(pl.UInt32))
+                    .otherwise(pl.col(val_col_name).list.len().cast(pl.UInt32))
+                    .alias(val_col_name)
+                )
+                [val_col_name]
+            )
+
         # ── Build the output DataFrame efficiently ────────────────────────
         # We need n_units × n_intervals rows in order:
         #   unit0×int0, unit0×int1, ..., unit0×intN, unit1×int0, ...
diff --git a/tests/test_join_between.py b/tests/test_join_between.py
index f896edb..d466f46 100644
--- a/tests/test_join_between.py
+++ b/tests/test_join_between.py
@@ -307,12 +307,12 @@ def test_join_between_as_counts_retains_columns(units, intervals):
 def test_join_between_vec_as_list_expr(units, intervals):
     """vec accepts arbitrary list expressions; clipping uses the transformed list.
 
-    Adding 1 to the list shifts all values up by 1, producing verifiably
-    different clipping results than the original column.
+    Adding 1 to each element of the list shifts all values up by 1, producing
+    verifiably different clipping results than the original column.
     """
     result = units.vec.join_between(
         other=intervals,
-        values=pl.col("event_times") + 1,
+        values=pl.col("event_times").list.eval(pl.element() + 1),
         bounds=("start_time", "stop_time"),
     ).sort("event_id", "interval_id")