Splice junctions are delayed by one position at the 5' side

[bkffadia]

I used regtools to extract splice junctions rom RNA-Seq data and noticed a delay by one position in the 5' side of the split read.
for exemple this splice jucntion in the positive strand, split read starts from 12665 to 13289, with the overhang being 32,69.

this suggests that the donor site would be at 12665 + 32 = 12697 and the donor site being at 13289-69=13220.
but when we view these sites at the UCSC genome browser it turns out that donor site is actually at 12698 :

whereas the acceptor site is correct :

for junctions in the negative strand :

this splice jucntion in the positive strand, split read starts from 25216 to 29348, with the overhang being 66,28.
this suggests that the donor site would be at 29348 - 28 = 29320 and the donor site being at 25216+66=25282.
but when we view both sites on the genome browser we noticed that the acceptor site being delayed by one position where as the donor site is correct :

so my question is, is this intentional or is it an error ?