Frequency for G correction

[gzentner]

As it stands now, I believe the frequency of G removal is calculated by rbinom using the frequency of reads with a soft-clipped G in all reads, if the +1 base is a G and there are no soft-clipped bases in a given read. Wouldn't an unencoded G be soft-clipped? I guess I'm just wondering why a read should have no soft-clipped bases to be considered for correction.

Relatedly, I think the frequency used for the binomial deviation should take into account the number of reads with a non-soft-clipped G at +1.