Whether the ORF region to be considered for primer designing should be between start and stop codon or some other parameter should be used to encapsulate maximum gene spanning region in annotated transcripts.
If those sequences came from a de novo assembly I would chose to design the primers on the UTR, as you most probably would amplify more than one mRNA if the primers are designed on the ORF.
If you have a reference genome, then go for the splice version you wanna see and design the primers on the ORF but spanning one intron. That would also get rid off the problem of gDNA contamination and your results would be more reliable.