German GOP - Low scores for first phoneme with stress

[Pranav Atre]

Hi, I am using Kaldi GOP recipe to find out phoneme wise goodness of pronunciation for German transcripts. I am using pre-trained model from here - https://github.com/uhh-lt/kaldi-tuda-de. 

In general, I am getting good results, but I am getting low scores for phonemes with primary stress (denoted by '). As seen in phones-pure file, it's using X-SAMPA notation.

I am attaching the gop results for the word 'Hallo' and align_lexocon file from lang dir  .

Can someone help me what exactly I need to do to increase the gop score of phonemes with stress notations?

align_lexicon.txt - https://drive.google.com/file/d/1LerrNWZtRw9qGEcqB0Zsin_-BEoddHoy/view?usp=sharing