pdf extraction has an issue while copying texts among vertically spanned cells

Describe the bug

I am extracting data from PDFs using camelot and am faced with the following issue on 3. page of this datasheet. The problematic table is shown below:

The issue is inconsistency during the copying content of span cells. As you can see on the following picture span cells are correctly detected.

Even if the cells are detected correctly in the 3. column the content is copied to one of two spanned cells and in the 4. column the content is copied to two of three spanned cells. You can see the data I extracted as follow. There is always one missing cell per both columns.

Steps to reproduce the bug

!pip install "camelot-py[cv]" -q
!pip install PyPDF2==2.12.1
!apt-get install ghostscript
import camelot
import pandas as pd
from tabulate import tabulate
import re
import fitz

Expected behavior

Code

table_areas=['86, 697, 529, 95'] # To ignore page borders
tables = camelot.read_pdf(single_source, pages='all', 
                          flavor = 'lattice', 
                          copy_text=['v'], 
                          line_scale = 110, 
                          table_regions=table_areas, 
                          flag_size = False, 
                          process_background=False)

to visualize the tables:

for table in accurate_tables:
  print(table.parsing_report, table.shape, table._bbox)
  print(tabulate(table.df, headers='keys', tablefmt='psql'))
  camelot.plot(table, kind='grid').show()

print("Extracting ", single_source, "is finished!")

PDF HERE

Screenshots

Environment

OS: MS and/or Google Colab
Python version: 3.8
Numpy version: last
OpenCV version: last
Ghostscript version: last
Camelot version: last
PyPDF2 version: <3.0

Additional context

camelot-dev / camelot

pdf extraction has an issue while copying texts among vertically spanned cells #349