minimap2 will generate MD tags. You could also use the second option to identify and the filter reads with perfect matches (if that is what you want).

--MD    Output the MD tag (see the SAM spec).
--eqx   Output =/X CIGAR operators for sequence match/mismatch.

While not a complete solution for your problem, this might help you further. The problem is that nanopore reads vary hugely in length, from tens of basepairs up to megabases, so mismatch filters must be dynamic relative to read length to make sense.

Samtools can help you out here if you have bam files. eg here to get those with less than 10% or 20% mismatches, just using the NM tag. You should be able to use this with other tags, eg MD.

The wc -l just gives you numbers - read counts, leave out to get the sequences in SAM format.

 # works with samtools 1.16.1 - must upgrade samtools from 1.11

samtools view -e '[NM]/length(seq) < 0.1' test.sorted.bam | wc -l
7
samtools view -e '[NM]/length(seq) < 0.2' test.sorted.bam | wc -l
12

BBTools has a tool for this purpose, "filtersam.sh". It works on bam too.

filtersam.sh in=mapped.sam out=filtered.sam ref=ref.fa

This will remove reads with multiple unsupported variations (the exact number and type is configurable). The goal was to remove junky Novaseq reads, where it's generally fine to discard all the reads with errors, since most reads don't have any... I'm not sure how good of an idea it is for Nanopore, though this data does look pretty clean.