Hi,
That is because the annotation data is transcript-centric and hence multiple (transcript ID) accessions per gene.
I do not know of any way to collapse the output to gene-level.
But, you could do that using Ensembl BioMart. Under the "Attributes" section (left-hand menu), for the "Features" section (right-hand main-panel), choose only Gene stable ID (and keep transcript stable ID un-checked). In the output section, check the "unique results only" option and the output would be gene-centric.
e.g. output showing coordinates at gene level