Invalid results in PGN files?

[HumbleProgrammer]

After adding logic to my PGN parser to compare the result (1-0, 0-1, 1/2-1/2) to the final board position, I am getting a much higher number of "failures" than I expected. I am using TWIC as the test corpus, and there are multiple games where the result and the final position don't match. The most common failure are games that are drawn by "repetition of position" yet adjudicated as a win for either player; this I understand, as the players may not recognize the repeated position at the time and resign or run out of time. The ones that I don't know how to handle are games where one player is marked as the winner, yet the board position shows them clearly checkmated!

I am seeking the collective wisdom of Open-Chess readers on how to handle this type of result mismatch:

(1) Do nothing. The PGN is always correct and there must have been a good reason at the time.

(2) Annotate the game with a comment pointing out the mismatch, but leave the result as originally posted.

(3) Correct the result, and add a comment narrating the change.

(4) Correct the result and move on.

Personally, I'm leaning towards (2) but would be very interested in feedback to the contrary.

Cheers!
Humble Programmer

[kingliveson]

(3) Correct the result, and add a comment narrating the change.