Transcriptome assembler comparisons-Getting vastly different results

Question

[4117837]

Hi there

I am comparing RNABLOOM2 (with and without Illumina shorts reads for polishing) and RATTLE for transcript assembly using reads from minion long read seq. The number of transcripts are extremely different. For RATTLE-50 000 transcripts; for RNABLOOM-about 160 000 transcripts (with and without short read polishing). Any idea why the vastly different transcript numbers? BUSCO scores are much better for the RNABLOOM (with short read polishing) assembly, though.

I thought maybe RATTLE collapses splice variants but it doesn't seem to, not based on their paper anyway.

Any help will be appreciated.