diff --git a/pdf4llm/setup.py b/pdf4llm/setup.py
index 79f2ccbf..c4af2cc0 100644
--- a/pdf4llm/setup.py
+++ b/pdf4llm/setup.py
@@ -6,7 +6,7 @@
 with open(os.path.join(setup_py_cwd, "README.md"), encoding="utf-8") as f:
     readme = f.read()
 
-version = "0.1.8"
+version = "0.1.9"
 
 classifiers = [
     "Development Status :: 5 - Production/Stable",
diff --git a/pymupdf4llm/pymupdf4llm/helpers/document_layout.py b/pymupdf4llm/pymupdf4llm/helpers/document_layout.py
index 34e1f392..c95377a7 100644
--- a/pymupdf4llm/pymupdf4llm/helpers/document_layout.py
+++ b/pymupdf4llm/pymupdf4llm/helpers/document_layout.py
@@ -8,10 +8,11 @@
 import pymupdf
 import tabulate
 from pymupdf4llm.helpers.get_text_lines import get_raw_lines
-from pymupdf4llm.helpers import utils, check_ocr
+from pymupdf4llm.helpers import utils
 
 try:
     import cv2
+    from pymupdf4llm.helpers import check_ocr
 except ImportError:
     cv2 = None
 
diff --git a/pymupdf4llm/pymupdf4llm/helpers/get_text_lines.py b/pymupdf4llm/pymupdf4llm/helpers/get_text_lines.py
index e1a02da9..f3ef2c94 100644
--- a/pymupdf4llm/pymupdf4llm/helpers/get_text_lines.py
+++ b/pymupdf4llm/pymupdf4llm/helpers/get_text_lines.py
@@ -93,8 +93,12 @@ def sanitize_spans(line):
             if s0["bbox"].x1 + delta < s1["bbox"].x0 or (
                 s0["flags"],
                 s0["char_flags"] & ~2,
-                s0["size"],
-            ) != (s1["flags"], s1["char_flags"] & ~2, s1["size"]):
+                # s0["size"],
+            ) != (
+                s1["flags"],
+                s1["char_flags"] & ~2,
+                # s1["size"],
+            ):
                 continue  # no joining
             # We need to join bbox and text of two consecutive spans
             # On occasion, spans may also be duplicated.
diff --git a/pymupdf4llm/pymupdf4llm/helpers/utils.py b/pymupdf4llm/pymupdf4llm/helpers/utils.py
index 28fea125..955e6046 100644
--- a/pymupdf4llm/pymupdf4llm/helpers/utils.py
+++ b/pymupdf4llm/pymupdf4llm/helpers/utils.py
@@ -221,11 +221,31 @@ def cluster_stripes(boxes, vertical_gap: float = 12):
     Returns:
         List of disjoint horizontal stripes. Each stripe is a list of boxes.
     """
+
+    def is_multi_column_layout(boxes):
+        sorted_boxes = sorted(boxes, key=lambda b: b[0])
+        columns = []
+        current_column = [sorted_boxes[0]]
+        for box in sorted_boxes[1:]:
+            prev_right = max([b[2] for b in current_column])
+            if box[0] - prev_right > 3:
+                columns.append(current_column)
+                current_column = [box]
+            else:
+                current_column.append(box)
+        columns.append(current_column)
+        return len(columns) > 1
+
     # Sort top to bottom
     sorted_boxes = sorted(boxes, key=lambda b: b[1])
     stripes = []
     if not sorted_boxes:
         return stripes
+
+    # Early exit for clean multi-column layouts
+    if is_multi_column_layout(sorted_boxes):
+        return [boxes]
+
     current_stripe = [sorted_boxes[0]]
 
     for box in sorted_boxes[1:]:
@@ -257,7 +277,7 @@ def cluster_columns_in_stripe(stripe: list):
 
     for box in sorted_boxes[1:]:
         prev_right = max([b[2] for b in current_column])
-        if box[0] - prev_right >= -1:
+        if box[0] - prev_right > 1:
             columns.append(sorted(current_column, key=lambda b: b[3]))
             current_column = [box]
         else:
@@ -292,14 +312,15 @@ def compute_reading_order(boxes, vertical_gap: float = 12):
     return ordered
 
 
-def find_reading_order(boxes, vertical_gap: float = 12) -> list:
+def find_reading_order(boxes, vertical_gap: float = 36) -> list:
     """Given page layout information, return the boxes in reading order.
 
     Args:
         boxes: List of classified bounding boxes with class info as defined
                by pymupdf_layout: (x0, y0, x1, y1, "class").
         vertical_gap: Minimum vertical gap to separate stripes. The default
-                      value of 12 works well for most documents.
+                      value of 36 works well for most documents. It roughly
+                      corresponds to 2 -3 text line heights
 
     Returns:
         List of boxes in reading order.
diff --git a/pymupdf4llm/pymupdf4llm/versions_file.py b/pymupdf4llm/pymupdf4llm/versions_file.py
index 7068efa9..cad78b1f 100644
--- a/pymupdf4llm/pymupdf4llm/versions_file.py
+++ b/pymupdf4llm/pymupdf4llm/versions_file.py
@@ -1,3 +1,3 @@
 # Generated file - do not edit.
 MINIMUM_PYMUPDF_VERSION = (1, 26, 6)
-VERSION = '0.1.8'
+VERSION = '0.1.9'
diff --git a/pymupdf4llm/setup.py b/pymupdf4llm/setup.py
index a788de42..7b329f1c 100644
--- a/pymupdf4llm/setup.py
+++ b/pymupdf4llm/setup.py
@@ -14,7 +14,7 @@
     "Topic :: Utilities",
 ]
 
-version = "0.1.8"
+version = "0.1.9"
 requires = ["pymupdf>=1.26.6", "tabulate"]
 
 text = requires[0].split("=")[1]