Update language model after successfully adding new words in small vosk models

Hi guys,

I was able to successfully add two new terms to the small english and german models following the instructions on [this site](https://chrisearch.wordpress.com/2017/03/11/speech-recognition-using-kaldi-extending-and-using-the-aspire-model/) (using utils/mkgraph_lookahead.sh instead to create dynamic graphs).
I replaced all the files in the graph directory of the original model with the newly created ones. Multiple test runs showed that the new words are recognized correctly.

However, now when I want to update the language model following the instructions on [this site](https://alphacephei.com/vosk/adaptation), the output is really bad and every word is recognized incorrectly. It should be mentioned that I already was able to do this update successfully before adding new words.

Am I doing anything wrong or is there something to consider that I don't know?