agglomerative method in snpgdsHCluster

[botanicpanic1]

I am using SNPRelate to produce a distance-based tree starting from a vcf file containing SNPs produced with ddRADseq.
I was curious to know which agglomerative method was employed in snpgdsHCluster and in the documentation is reported that: "Call the function hclust to perform hierarchical cluster analysis, using method="average"". So basically snpgdsHCluster uses UPGMA. But usually tree crafted using UPGMA should be ultrametric, however the output is an additive tree. Is there an explanation for this? Thanks to whoever will respond!