diff --git a/tasktiger/task.py b/tasktiger/task.py
index 2343ba60..d5b5bb5b 100644
--- a/tasktiger/task.py
+++ b/tasktiger/task.py
@@ -513,15 +513,19 @@ def tasks_from_queue(
             tiger: TaskTiger instance.
             queue: Name of the queue.
             state: State of the task (QUEUED, ACTIVE, SCHEDULED, ERROR).
+            skip: Number of latest tasks to pass over.
             limit: Maximum number of tasks to return.
             load_executions: Maximum number of executions to load for each task
                 (starting from the latest).
-            include_not_found: Whether to include tasks that cannot be loaded.
+            include_not_found: Whether to include tasks that have no payload.
 
         Returns:
             Tuple with the following information:
-            * total items in the queue
-            * tasks from the given queue in the given state, latest first.
+            * total items in the queue, including tasks without a payload
+            * the latest `limit` tasks after the latest `skip` tasks, in
+              ascending order of score. If tasks without a payload are left
+              out, this list can be shorter than `limit` before the end of the
+              queue.
         """
 
         key = tiger._key(state, queue)
@@ -549,8 +553,9 @@ def tasks_from_queue(
                     range(len(items)), results[0], results[1:], tss
                 ):
                     if serialized_data is None:
-                        if include_not_found:
-                            data = {"id": items[idx][0]}
+                        if not include_not_found:
+                            continue
+                        data = {"id": items[idx][0]}
                     else:
                         data = json.loads(serialized_data)
 
@@ -572,8 +577,9 @@ def tasks_from_queue(
                 )
                 for idx, serialized_data, ts in zip(range(len(items)), result, tss):
                     if serialized_data is None:
-                        if include_not_found:
-                            data = {"id": items[idx][0]}
+                        if not include_not_found:
+                            continue
+                        data = {"id": items[idx][0]}
                     else:
                         data = json.loads(serialized_data)
 
diff --git a/tests/test_base.py b/tests/test_base.py
index df05ea1c..26c7a79b 100644
--- a/tests/test_base.py
+++ b/tests/test_base.py
@@ -1072,6 +1072,35 @@ def test_tasks_from_queue_with_executions(self):
         assert n == 1
         assert len(tasks[0].executions) == 2
 
+    @pytest.mark.parametrize("load_executions", [0, 1])
+    @pytest.mark.parametrize("missing_index", [0, 1, 2])
+    def test_tasks_from_queue_task_not_found(self, load_executions, missing_index):
+        tasks = [Task(self.tiger, simple_task) for _ in range(3)]
+        # Distinct scheduled times give a fixed order in the zset.
+        for minutes, task in enumerate(tasks, start=1):
+            task.delay(when=datetime.timedelta(minutes=minutes))
+        missing_task = tasks[missing_index]
+        self.conn.delete(self.tiger._key("task", missing_task.id))
+
+        n, found = Task.tasks_from_queue(
+            self.tiger, "default", "scheduled", load_executions=load_executions
+        )
+        assert n == 3
+        assert [task.id for task in found] == [
+            task.id for task in tasks if task is not missing_task
+        ]
+
+        n, found = Task.tasks_from_queue(
+            self.tiger,
+            "default",
+            "scheduled",
+            load_executions=load_executions,
+            include_not_found=True,
+        )
+        assert n == 3
+        assert [task.id for task in found] == [task.id for task in tasks]
+        assert found[missing_index].data == {"id": missing_task.id}
+
     def test_eager(self):
         self.tiger.config["ALWAYS_EAGER"] = True