samtools mpileup for SNP genotyping (VCF4)

[kkoh]

Hi all,
Background
I have six samples (s1-s6), the breakdown of the SNP (ref/alt) at contig1 position 3929 is as follows:-

sample,ref_count,alt_count
s1, 3, 0
s2, 0, 16
s3, 24, 0
s4, 1, 16
s5, 14, 0
s6, 8, 14

Most (if not all) called bases have reasonable quality scores.

Synopsis
I am trying to understand the output generated by samtools mpileup ( 0.1.12-10 (r896), default parameters) below

----
contig1 3929 . G T 6.9 . DP=96;AF1=0.2036;
CI95=0.1667,0.3333;DP4=36,14,5,0;MQ=20;PV4=0.31,4.6e-18,1,0.11
PL:GT:GQ
0,9,56:0/0:12
30,6,0:0/1:5
0,72,142:0/0:75
25,0,11:0/1:18
0,42,154:0/0:45
0,10,106:0/0:13
----
Some strange observations :
a) the QUAL column=6.9, which indicates Phred-scaled probability of all samples being homozygous reference, seems unreasonably low, given the number of reads that support "ALT";
b) one would think GQ (genotype quality) for s2 should be a lot higher,given that there is zero count from REF;
c) shouldn't “0/1” (instead of 0/0) make more sense for s6.

Did I miss something here? Appreciate any help I can get.