Script reads cells of files row by row and tries to understand what rows are same on the same place, what rows are same but were moved, what rows were modified and what rows were deleted or added. Of course it is possible that a row has lots of modified cells and in this case script will consider that it old one was removed and new one was added. To reduce probability of this case you can enter 'valuable columns' and specify the rows that are 'header' of the table.