GATK can recalibrate the quality scores in a sequencing run. However, would amplicon sequencing produce a rich enough data set to train the recalibrater.
For example, pretend I have one amplicon in my sequencing run with a novel SNP at a given position. This position would presumably get marked as low quality (since all the reads disagree with the reference). This means that trying to call SNPs from recalibrated scores would produce incorrect qualities, especially at variant sites?