Hello @AgustinGV
I’m not exactly sure what you are trying to do but that you can explain more if needed.
STAR will not create a reference assembly.
You can use an existing assembly from your target genome, in whole or in part, as a Custom Genome/Transcriptome to call variants that exist in the RNA-seq dataset. You can also assemble a reference transcriptome.
That said, I would suggest reviewing the publication related to the work you are trying to replicate – what reference did they use? It should be the same in your work or expect different results/statistics.
Please see this related prior Q&A for more options. The tutorials under “Variant Analysis” would also probably apply for your case. Galaxy tools for the de novo assembled transcriptome annotation
Thanks!