The 13k concepts I said is got from the pandas dataframe of:
df = umls.map_umls2icd10()
It seems there are only 13k concepts in the umls files that have a icd 10 code assigned to it.
Also, when i checked the official demo of medcat I got icd 10 codes for concepts that had no icd 10 codes in my umls to ics 10 mapped model. Does the demo use a different model? If that is the case that model can solve my problem.