I think it could be done using a two-step approach.

1. First extract all the reads that are mapped within your region of interest using samtools view command.

   samtools view Input.bam chrA:x-y | cut -f1 > IDs.txt
   

2. Now you can use these read ids and extract both forward and reverse reads corresponding to that read id from the original bam file. You can use a single grep command with multiple strings or read ids. See egrep and |. Of course , this is not an efficient way but it should work.

egrep "^#|Id1|Id2|....|Idn" Input.sam (sam format) > Output.sam or you can use grep -f IDs.txt Input.sam and add header afterwards.

I wrote a detailed example on how to do that and you can run the code and go through the details in the code comments here : http://coderscrowd.com/app/codes/view/244 for your example I would do it on the mates output the results in bed files and use bedtools to remap the results Hope that helps

I wrote something today: http://lindenb.github.io/jvarkit/SamViewWithMate.html

$ java -jar dist/samviewwithmate.jar -r "9:137230721-137230796"  ./src/test/resources/HG02260.transloc.chr9.14.bam | cut -f 1-9 | tail
ERR251239.10989793  83  9   137230747   60  30S70M  =   137230326   -490
ERR251239.3385449   147 9   137230754   60  1S99M   =   137230352   -500
ERR251240.17111373  99  9   137230764   60  100M    =   137231150   475
ERR251240.46859433  147 9   137230777   60  65S35M  =   137230342   -469
ERR251240.74563730  147 9   137230787   60  1S99M   =   137230407   -478
ERR251240.1291708   83  9   137230789   60  100M    =   137230411   -477
ERR251240.11887757  97  9   137230795   37  100M    14  79839451    0
ERR251239.34016218  81  14  79839349    37  100M    9   137230679   0
ERR251240.10196873  81  14  79839368    37  100M    9   137230721   0
ERR251240.11887757  145 14  79839451    37  100M    9   137230795   0