I am trying to slice some whole exome bam files from TCGA and for my pipeline I need to extract read pairs. I used samtools sort -n to sort the bam file and using bedtools bamtofastq to extract paired reads but it seems there are lot of reads for which there is no mate in the bam file. I thought I could extract the fastq files from the bam files which will be equivalent to raw sequencing data but it does not seems to be the case. Please share your experience or if you have better method to extract reads. I am going to try Sambamba which some folks suggest to be much faster than samtools.