--no-trimming leads to significant drop in BUSCO scores

[JWDebler]

Hi,
I just came accross something strange when comparing my genome after different rounds and settings of polishing.

I assembled Q20 Nanopore reads with Canu and then polished once with racon with or without the --no-trimming option, because in the past I have noticed that telomeres get removed without the --no-trimming option.
But this time the problem seems different. When the Canu assembly gets polished with the --no-trimming option, the assembly actually seems to be getting worse.

Going back over some 'old' assemblies from ONT R9 series flowcell and chemistry I see the same thing. Only happens when --no-trimming is used though. I always run Medaka after racon which seems to 'rescue' whatever racon was doing, but this makes me wonder if I should skip racon completely in the future.

My full command is:
minimap2 assembly.fasta reads.fastq.gz > minimap.racon.paf

racon -m 8 -x -6 -g -8 -w 500 -t 14 --no-trimming reads.fastq.gz minimap.racon.paf assembly.fasta > assembly_racon.fasta

I have tried this with the latest version 1.5.0 as well as the previous 1.4.21.