These reads all come from the same sequence in the reference.

test_sort_1: Phred scores around 10 (worst), exact match for reference (best)
test_sort_2: Phred scores around 20 (best), error rate of about 3% (worst)
test_sort_3: Phred scores around 15 (medium), error rate of about 1% (medium)

Quality order by Phred:     2, 3, 1
Quality order by reference: 1, 3, 2