diff --git a/src/calibre/ebooks/mobi/debug.py b/src/calibre/ebooks/mobi/debug.py
index 12bdb41f4b..1279ba7793 100644
--- a/src/calibre/ebooks/mobi/debug.py
+++ b/src/calibre/ebooks/mobi/debug.py
@@ -844,6 +844,7 @@ class TextRecord(object): # {{{
 
     def __init__(self, idx, record, extra_data_flags, decompress):
         self.trailing_data, self.raw = get_trailing_data(record.raw, extra_data_flags)
+        raw_trailing_bytes = record.raw[len(self.raw):]
         self.raw = decompress(self.raw)
         if 0 in self.trailing_data:
             self.trailing_data['multibyte_overlap'] = self.trailing_data.pop(0)
@@ -851,6 +852,7 @@ class TextRecord(object): # {{{
             self.trailing_data['indexing'] = self.trailing_data.pop(1)
         if 2 in self.trailing_data:
             self.trailing_data['uncrossable_breaks'] = self.trailing_data.pop(2)
+        self.trailing_data['raw_bytes'] = raw_trailing_bytes
 
         self.idx = idx
 
diff --git a/src/calibre/ebooks/mobi/utils.py b/src/calibre/ebooks/mobi/utils.py
index 839374af70..6df9db3b3b 100644
--- a/src/calibre/ebooks/mobi/utils.py
+++ b/src/calibre/ebooks/mobi/utils.py
@@ -191,7 +191,7 @@ def encode_trailing_data(raw):
         <data><size>
 
     where size is a backwards encoded vwi whose value is the length of the
-    entire return bytestring.
+    entire returned bytestring. data is the bytestring passed in as raw.
 
     This is the encoding used for trailing data entries at the end of text
     records. See get_trailing_data() for details.
diff --git a/src/calibre/ebooks/mobi/writer2/indexer.py b/src/calibre/ebooks/mobi/writer2/indexer.py
index 14c5328622..f121e29835 100644
--- a/src/calibre/ebooks/mobi/writer2/indexer.py
+++ b/src/calibre/ebooks/mobi/writer2/indexer.py
@@ -14,7 +14,7 @@ from collections import OrderedDict, defaultdict
 
 from calibre.ebooks.mobi.writer2 import RECORD_SIZE
 from calibre.ebooks.mobi.utils import (encint, encode_number_as_hex,
-        encode_trailing_data, encode_tbs, align_block, utf8_text)
+        encode_tbs, align_block, utf8_text)
 
 
 class CNCX(object): # {{{
@@ -198,7 +198,7 @@ class TBS(object): # {{{
                     # This can happen if a record contains only text between
                     # the periodical start and the first section
                     byts = self.type_011
-                self.bytestring = encode_trailing_data(byts)
+                self.bytestring = byts
             else:
                 depth_map = defaultdict(list)
                 for x in ('starts', 'ends', 'completes'):
@@ -209,7 +209,7 @@ class TBS(object): # {{{
                 self.periodical_tbs(data, first, depth_map)
         else:
             if not data:
-                self.bytestring = encode_trailing_data(b'')
+                self.bytestring = b''
             else:
                 self.book_tbs(data, first)
 
@@ -302,10 +302,10 @@ class TBS(object): # {{{
             buf.write(encode_tbs(spanner.index - parent_section_index,
                 {0b0001: 0}))
 
-        self.bytestring = encode_trailing_data(buf.getvalue())
+        self.bytestring = buf.getvalue()
 
     def book_tbs(self, data, first):
-        self.bytestring = encode_trailing_data(b'')
+        self.bytestring = b''
 # }}}
 
 class Indexer(object): # {{{
diff --git a/src/calibre/ebooks/mobi/writer2/main.py b/src/calibre/ebooks/mobi/writer2/main.py
index e13afa2ba7..44c471d3d4 100644
--- a/src/calibre/ebooks/mobi/writer2/main.py
+++ b/src/calibre/ebooks/mobi/writer2/main.py
@@ -99,7 +99,7 @@ class MobiWriter(object):
             for i in xrange(len(self.records)):
                 if i == 0: continue
                 tbs = self.indexer.get_trailing_byte_sequence(i)
-                self.records[i] += tbs
+                self.records[i] += encode_trailing_data(tbs)
             self.records.extend(self.indexer.records)
 
     @property
@@ -212,15 +212,15 @@ class MobiWriter(object):
             if self.compression == PALMDOC:
                 data = compress_doc(data)
             record = StringIO()
-            record.write(data)
 
-            self.records.append(record.getvalue())
             nrecords += 1
             data, overlap = self.read_text_record(text)
+            record.write(data)
 
-            # Write information about the mutibyte character overlap, if any
+            # Write information about the multibyte character overlap, if any
             record.write(overlap)
             record.write(pack(b'>B', len(overlap)))
+            self.records.append(record.getvalue())
 
         self.last_text_record_idx = nrecords