Handle Not Applicable (NA) response in Likert scale

Hi All,
I measured the performance of field Trainers through questions using Likert scale (0 to 5).
0 will be given for the poor performance. 5 for the best performance. On this basis, each trainer will get some marks out of maximum.
For Eg : We have total of 5 Questions and each question has 5 likert scale . So the maximum marks a trainer can get is 25 and according to their performance they get categories in different level.

But there are some items/questions that are not relevant for the trainer and we cannot assess him/her on that item, then in this case, I have given one more option Not Applicable (NA)/Can't Say response apart from the 5 point scale.

But when I am analyzing the data, I am not able to understand how should i code these (Not Applicable (NA)/Can't Say) response and analyse them with other quantitative data.

What would be the best approach to treat the NA response? If I code them as some other value like (9 or 99) then it can increase my maximum value.

Please suggest best approach.

Reagrds,
Deeksha sharma