RNAseq preferred reference - b37 or hg19?

Hi. I'm wondering if there's a preferred human reference build for aligning RNAseq data?

The only factor that even leads me to ask this question is the following article
https://software.broadinstitute.org/gatk/documentation/article.php?id=3891
on RNAseq best practices, in which hg19 is used. And I gather that b37 is pretty much ubiquitously used at the Broad Institute, so I just wanted to make sure.

(I suspect it makes no difference - sorry for the nit-picking.)

Thanks,
Jeff