Why does Txdb transcript length not always match to transcript end-start position?

I have just found an example that biomart’s transcript_length is not identical with transcript_end – transcript_start.

ensembl_gene_id mgi_symbol chromosome_name strand start_position end_position   gene_biotype transcript_start transcript_end strand.1 transcript_length
128537 ENSMUSG00000037860       Aim2               1      1      173178445    173293606 protein_coding        173178445      173293606        1              2839
128538 ENSMUSG00000037860       Aim2               1      1      173178445    173293606 protein_coding        173246762      173255347        1               383
128539 ENSMUSG00000037860       Aim2               1      1      173178445    173293606 protein_coding        173248164      173287285        1               744

Does anyone know why such discrepancy happen??

Thanks