problem with transcript to genomic coordinates conversion when UTR spans more than 1 exon

[XiaoJianfeng]

transvar canno --refseq -i 'NM_000051.3:c.-30' --refversion hg19
input transcript gene strand coordinates(gDNA/cDNA/protein) region info
NM_000051.3:c.-30 NM_000051.3 (protein_coding) ATM + chr11:g.108098322A/c.1-30A/. inside_[5-UTR;noncoding_exon_2] C2=NextToSpliceAcceptorOfExon2_At_chr11:108098321;dbxref=GeneID:472,HGNC:795,MIM:607585;aliases=NP_000042;source=RefSeq
transvar canno --refseq -i 'NM_000051.3:c.-31' --refversion hg19
input transcript gene strand coordinates(gDNA/cDNA/protein) region info
NM_000051.3:c.-31 NM_000051.3 (protein_coding) ATM + chr11:g.108098321G/c.1-31G/. inside_[5-UTR;intron_between_exon_1_and_2] C2=SpliceAcceptorOfExon2_At_chr11:108098321;dbxref=GeneID:472,HGNC:795,MIM:607585;aliases=NP_000042;source=RefSeq

The position transvar gets for c.-31 is chr11:g.108098321. In fact, there is a large intron between c.-30 (the left most base of the exon 2) and c.-31 (the right most base of exon 1), and they are not next to each other. It seems transvar assumes UTR is always on the 1st exon.