diff --git a/recce/adapter/dbt_adapter/__init__.py b/recce/adapter/dbt_adapter/__init__.py
index e31ce3c4..d6d74123 100644
--- a/recce/adapter/dbt_adapter/__init__.py
+++ b/recce/adapter/dbt_adapter/__init__.py
@@ -631,13 +631,37 @@ def select_nodes(self, select: Optional[str] = None, exclude: Optional[str] = No
             spec = parse_difference(select_list, exclude_list, "eager")
         else:
             spec = parse_difference(select_list, exclude_list)
+
+        manifest = self.manifest.deepcopy()
+
+        for (key, node) in self.previous_state.manifest.nodes.items():
+            if key not in manifest.nodes:
+                manifest.nodes[key] = node
+
+        for (key, node) in self.previous_state.manifest.sources.items():
+            if key not in manifest.sources:
+                manifest.sources[key] = node
+
+        for (key, node) in self.previous_state.manifest.exposures.items():
+            if key not in manifest.exposures:
+                manifest.exposures[key] = node
+
+        for (key, node) in self.previous_state.manifest.metrics.items():
+            if key not in manifest.metrics:
+                manifest.metrics[key] = node
+
+        if hasattr(self.previous_state.manifest, 'semantic_models'):
+            for (key, node) in self.previous_state.manifest.semantic_models.items():
+                if key not in manifest.semantic_models:
+                    manifest.semantic_models[key] = node
+
         compiler = Compiler(self.runtime_config)
         # disable to print compile states
         tmp_func = dbt.compilation.print_compile_stats
         dbt.compilation.print_compile_stats = lambda x: None
-        graph = compiler.compile(self.manifest, write=False)
+        graph = compiler.compile(manifest, write=False)
         dbt.compilation.print_compile_stats = tmp_func
-        selector = NodeSelector(graph, self.manifest, previous_state=self.previous_state)
+        selector = NodeSelector(graph, manifest, previous_state=self.previous_state)
 
         return selector.get_selected(spec)
 
diff --git a/tests/adapter/dbt_adapter/dbt_test_helper.py b/tests/adapter/dbt_adapter/dbt_test_helper.py
index 6f04f376..39cf8d2b 100644
--- a/tests/adapter/dbt_adapter/dbt_test_helper.py
+++ b/tests/adapter/dbt_adapter/dbt_test_helper.py
@@ -41,7 +41,7 @@ def __init__(self):
         self.adapter.execute(f"CREATE schema IF NOT EXISTS {self.curr_schema}")
         self.adapter.set_artifacts(self.base_manifest, self.curr_manifest)
 
-    def create_model(self, model_name, base_csv, curr_csv, depends_on=[]):
+    def create_model(self, model_name, base_csv=None, curr_csv=None, depends_on=[]):
         package_name = "recce_test"
         # unique_id = f"model.{package_name}.{model_name}"
         unique_id = model_name
@@ -84,19 +84,19 @@ def _add_model_to_manifest(base, raw_code):
             manifest.add_node_nofile(node)
             return node
 
-        base_csv = textwrap.dedent(base_csv)
-        curr_csv = textwrap.dedent(curr_csv)
-
-        _add_model_to_manifest(True, base_csv)
-        _add_model_to_manifest(False, curr_csv)
-
-        import pandas as pd
-        df_base = pd.read_csv(StringIO(base_csv))
-        df_curr = pd.read_csv(StringIO(curr_csv))
         dbt_adapter = self.adapter
         with dbt_adapter.connection_named('create model'):
-            dbt_adapter.execute(f"CREATE TABLE {self.base_schema}.{model_name} AS SELECT * FROM df_base")
-            dbt_adapter.execute(f"CREATE TABLE {self.curr_schema}.{model_name} AS SELECT * FROM df_curr")
+            import pandas as pd
+            if base_csv:
+                base_csv = textwrap.dedent(base_csv)
+                _add_model_to_manifest(True, base_csv)
+                df_base = pd.read_csv(StringIO(base_csv))
+                dbt_adapter.execute(f"CREATE TABLE {self.base_schema}.{model_name} AS SELECT * FROM df_base")
+            if curr_csv:
+                curr_csv = textwrap.dedent(curr_csv)
+                _add_model_to_manifest(False, curr_csv)
+                df_curr = pd.read_csv(StringIO(curr_csv))
+                dbt_adapter.execute(f"CREATE TABLE {self.curr_schema}.{model_name} AS SELECT * FROM df_curr")
         self.adapter.set_artifacts(self.base_manifest, self.curr_manifest)
 
     def remove_model(self, model_name):
diff --git a/tests/adapter/dbt_adapter/test_selector.py b/tests/adapter/dbt_adapter/test_selector.py
index 08463fce..5703d636 100644
--- a/tests/adapter/dbt_adapter/test_selector.py
+++ b/tests/adapter/dbt_adapter/test_selector.py
@@ -70,3 +70,32 @@ def test_select(dbt_test_helper):
 
     node_ids = adapter.select_nodes("state:modified,resource_type:snapshot")
     assert len(node_ids) == 1
+
+
+def test_select_removed_by_graph(dbt_test_helper):
+    csv_data_curr = """
+        customer_id,name,age
+        1,Alice,30
+        2,Bob,25
+        3,Charlie,35
+        """
+
+    csv_data_base = """
+        customer_id,name,age
+        1,Alice,35
+        2,Bob,25
+        3,Charlie,35
+        """
+
+    dbt_test_helper.create_model("customers_1", csv_data_base, csv_data_curr)
+    dbt_test_helper.create_model("customers_2", base_csv=csv_data_base, depends_on=["customers_1"])
+    dbt_test_helper.create_model("customers_3", curr_csv=csv_data_curr, depends_on=["customers_1"])
+    adapter: DbtAdapter = dbt_test_helper.context.adapter
+
+    # Test graph operation
+    node_ids = adapter.select_nodes("customers_1+")
+    assert len(node_ids) == 3
+    node_ids = adapter.select_nodes("+customers_2")
+    assert len(node_ids) == 2
+    node_ids = adapter.select_nodes("+customers_3")
+    assert len(node_ids) == 2