I think the unfortunate thing about the current version of this nf-core pipeline is that the better performing tools are not reliable to run and ribotricer was chosen as it was easy to run rather than it being the best tool. There are no clear cut winners though and in reality translated region calling cannot be reliably done with samples with normal sequencing depth and annotation should be done on large amounts of aggregated data. I would argue that quantifying the expression of known non-CDS regions is more in the remit of this pipeline than translated region annotation would be.