The problem is probably the text file. But it's hard to know without seeing it.

It would be helpful to have a small patch and a text file demonstrating the problem.

The short answer is that there are many ways of encoding text in the world. About the only thing there is agreement on is the non-accented Latin alphabet, plus digits and a handful of punctuation marks (and even there, there has been disagreement). There is enough standardization now that it ought to be possible to take a text file generated on any computer and operating system and use it on another computer and operating system. But sometimes there needs to be some special handling. Without a concrete example, however, it's impossible to say what that handling needs to be.