Strand inconsistency between Long read tool and minimap2?

Question

[shinyjj]

I found there is an inconsistency on LR strand between LR gene assignment table by IsoQuant and LR alignment bam file by minimap. Do you know what could cause this inconsistency? I see this NM_006988.5 is - strand according to refseq annotation so IsoQuant gave it correctly but not sure why minimap2 is giving + strand. I gave refseq to IsoQuant when running and used the following command for minimap2. I also gave junctions_bed which gives where splicing junctions are to be more precise. Is this because long read is error-prone?

./minimap2 -a -Y --junc-bed <junctions.bed> -k 14 -x splice -t 20