diff --git a/community/kernel/src/main/java/org/neo4j/kernel/impl/index/schema/NativeSchemaIndexReader.java b/community/kernel/src/main/java/org/neo4j/kernel/impl/index/schema/NativeSchemaIndexReader.java
index c15c404c40208..611365ddd541a 100644
--- a/community/kernel/src/main/java/org/neo4j/kernel/impl/index/schema/NativeSchemaIndexReader.java
+++ b/community/kernel/src/main/java/org/neo4j/kernel/impl/index/schema/NativeSchemaIndexReader.java
@@ -70,8 +70,8 @@ public void close()
     @Override
     public IndexSampler createSampler()
     {
-        // For an unique index there's an optimization, knowing that all values in it are unique, to simply count
-        // the number of indexes values and create a sample for that count. The GBPTree doesn't have an O(1)
+        // For a unique index there's an optimization, knowing that all values in it are unique, to simply count
+        // the number of indexed values and create a sample for that count. The GBPTree doesn't have an O(1)
         // count mechanism, it will have to manually count the indexed values in it to get it.
         // For that reason this implementation opts for keeping complexity down by just using the existing
         // non-unique sampler which scans the index and counts (potentially duplicates, of which there will
diff --git a/community/neo4j/src/test/java/org/neo4j/index/IndexSamplingIntegrationTest.java b/community/neo4j/src/test/java/org/neo4j/index/IndexSamplingIntegrationTest.java
index c5b1849e6bfbf..60f5e46802023 100644
--- a/community/neo4j/src/test/java/org/neo4j/index/IndexSamplingIntegrationTest.java
+++ b/community/neo4j/src/test/java/org/neo4j/index/IndexSamplingIntegrationTest.java
@@ -51,7 +51,11 @@
 import org.neo4j.test.TestGraphDatabaseFactory;
 import org.neo4j.test.rule.TestDirectory;
 
+import static org.hamcrest.Matchers.allOf;
+import static org.hamcrest.Matchers.greaterThanOrEqualTo;
+import static org.hamcrest.Matchers.lessThanOrEqualTo;
 import static org.junit.Assert.assertEquals;
+import static org.junit.Assert.assertThat;
 
 public class IndexSamplingIntegrationTest
 {
@@ -60,7 +64,7 @@ public class IndexSamplingIntegrationTest
 
     private final Label label = Label.label( "Person" );
     private final String property = "name";
-    private final int nodes = 1000;
+    private final long nodes = 1000;
     private final String[] names = {"Neo4j", "Neo", "Graph", "Apa"};
 
     @Test
@@ -121,12 +125,12 @@ public void shouldSampleNotUniqueIndex() throws Throwable
 
         // Then
 
-        // sampling will consider also the delete nodes till the next lucene compaction
+        // lucene will consider also the delete nodes, native won't
         DoubleLongRegister register = fetchIndexSamplingValues( db );
         assertEquals( names.length, register.readFirst() );
-        assertEquals( nodes, register.readSecond() );
+        assertThat( register.readSecond(), allOf( greaterThanOrEqualTo( nodes - deletedNodes ), lessThanOrEqualTo( nodes ) ) );
 
-        // but the deleted nodes should not be considered in the index size value
+        // but regardless, the deleted nodes should not be considered in the index size value
         DoubleLongRegister indexSizeRegister = fetchIndexSizeValues( db );
         assertEquals( 0, indexSizeRegister.readFirst() );
         assertEquals( nodes - deletedNodes, indexSizeRegister.readSecond() );