I've attached an example of EPHA2, which has the name MSTRG.59 in the gene records.
Also attached are a report on all of the gene name mismatches
gff3 is at https://s3-us-west-2.amazonaws.com/human-pangenomics/T2T/CHM13/assemblies/annotation/chm13.draft_v2.0.gene_annotation.gff3
I suggest adding a sanity check before writing the file that all of the records for a given gene have the same gene_name
MSTRG.59-EPHA2.gff3.gz
gene_annotation_probs.tsv.gz