On using longer RNA-seq reads to improve transcript prediction accuracy