Should I run hisat2 in a 2-pass mode for better mapping at novel junctions?

[weishwu]

Is it recommended to run HISAT2 in a 2-pass mode in which the first run is to generate the splice junctions, and then the 2nd run is to use those junctions to align the reads again? STAR has a 2-pass mode so I wonder what is the equivalent way to run HISAT2. The manual says that, by default HISAT2 "makes use of splice sites found by earlier reads to align later reads in the same run". Does this mean pass 2 is actually not necessary?
My goal is transcriptome assembly using Stringtie, so accurate mapping at novel junctions is preferred.