diff --git a/src/bitmap/utils/chunk_iterator/mod.rs b/src/bitmap/utils/chunk_iterator/mod.rs
index 26df6ee8273..b896b1a6662 100644
--- a/src/bitmap/utils/chunk_iterator/mod.rs
+++ b/src/bitmap/utils/chunk_iterator/mod.rs
@@ -31,11 +31,9 @@ pub struct BitChunks<'a, T: BitChunk> {
 /// writes `bytes` into `dst`.
 #[inline]
 fn copy_with_merge<T: BitChunk>(dst: &mut T::Bytes, bytes: &[u8], bit_offset: usize) {
-    let mut last = bytes[bytes.len() - 1];
-    last >>= bit_offset;
-    dst[0] = last;
     bytes
         .windows(2)
+        .chain(std::iter::once([bytes[bytes.len() - 1], 0].as_ref()))
         .take(std::mem::size_of::<T>())
         .enumerate()
         .for_each(|(i, w)| {
@@ -340,4 +338,23 @@ mod tests {
         assert_eq!(iter.next(), None);
         assert_eq!(iter.remainder(), 0b1_1111_1111u64);
     }
+
+    #[test]
+    fn remainder_2() {
+        // (i % 3 == 0) in bitmap
+        let input: &[u8] = &[
+            0b01001001, 0b10010010, 0b00100100, 0b01001001, 0b10010010, 0b00100100, 0b01001001,
+            0b10010010, 0b00100100, 0b01001001, /* 73 */
+            0b10010010, /* 146 */
+            0b00100100, 0b00001001,
+        ];
+        let offset = 10; // 8 + 2
+        let length = 90;
+
+        let mut iter = BitChunks::<u64>::new(input, offset, length);
+        let first: u64 = 0b0100100100100100100100100100100100100100100100100100100100100100;
+        assert_eq!(first, iter.next().unwrap());
+        assert_eq!(iter.next(), None);
+        assert_eq!(iter.remainder(), 0b10010010010010010010010010u64);
+    }
 }
diff --git a/src/bitmap/utils/mod.rs b/src/bitmap/utils/mod.rs
index db284b28d76..538c193b0b2 100644
--- a/src/bitmap/utils/mod.rs
+++ b/src/bitmap/utils/mod.rs
@@ -65,6 +65,8 @@ pub fn bytes_for(bits: usize) -> usize {
 
 #[inline]
 pub fn null_count(slice: &[u8], offset: usize, len: usize) -> usize {
+    //return BitmapIter::new(slice, offset, len).filter(|x| !*x).count();
+
     // u64 results in optimal performance (verified via benches)
     let mut chunks = chunk_iterator::BitChunks::<u64>::new(slice, offset, len);
 
@@ -138,4 +140,11 @@ mod tests {
         assert_eq!(null_count(input, 13, 2), 1);
         assert_eq!(null_count(input, 14, 2), 1);
     }
+
+    #[test]
+    fn null_count_1() {
+        // offset = 10, len = 90 => remainder
+        let input: &[u8] = &[73, 146, 36, 73, 146, 36, 73, 146, 36, 73, 146, 36, 9];
+        assert_eq!(null_count(input, 10, 90), 60);
+    }
 }