Interpreting Local Realignment results from GATK

[alma]

Hi there. I need some help in interpreting the result of Local Realignment (around indels) using GATK please. I am new to using the package and trying to understand it. I have run the local realignment around indels according to the GATK manual to reduce the rate of mismatched alignment and false positive SNPs. After running the suspicious-interval-finding step, the output file gives the following 3 messages:
-2757553 reads were filtered out during traversal out of 73748806 total (3.74%)
- 639389 reads (0.87% of total) failing ZeroMappingQualityReadFilter
- 2118164 reads (2.87% of total) failing UnmappedReadFilter
Now here are the problems/questions:
Q1. Word counting (wc –l) of my input.bam file shows that I have 96,378,717 number of reads. Why the programme reports only 73,748,806 reads in the first message?
Q2. Similarly, the number of reads with zero MAPQ in the input.bam file is 25,809,346 whereas the output says 639,389 in the 2nd message. Similarly my count of the number of unmapped read is 23,055,530 which is much higher compared to that reported in the 3rd messages (2,118,164). Why these anomalies?
Q4. After realignment I get the message:
“0 reads were filtered out during traversal out of 12,369,995 total (0.00%)”
Word count ( wc –l) of realigned bam file shows that there are 12,369,991 number of reads, which is almost same as reported. But I expected to see at least 70,991,253 (73,748,806 −2,757,553) number of reads in the realigned bam file considering the read numbers reported in first message are correct. Such a low number of reads in the realigned bam file compared to input bam file is a real concern. Can anyone suggest what is going on?

alma