diff --git a/src/python/preproc_docs.py b/src/python/preproc_docs.py
index 70946ac229bca9066637902a35cc80becea5082b..6b4b805a916cc8c42a0b11ac1c0b9f65ffe78b7f 100644
--- a/src/python/preproc_docs.py
+++ b/src/python/preproc_docs.py
@@ -951,6 +951,7 @@ def group_textl_create_xml(xml_t, set_of_blocks, ref_textl, labels_textl, rescal
             all_bbox = np.array([]).reshape((4,0))
 
 
+    # TODO note clemens: this might be the reason there are still textlines with raw character children at the end of the document
     # Just add the two final elements from the original xml
     page_el.append(xml_t[0][-2]) # Figure
     page_el.append(xml_t[0][-2]) # Layout