bwa mem -a and -t options are incompatible?

[yazhinia]

Hi,
When I wanted to run bwa mem with all alignment options -a and parallelization using -t, I find that the output only contains the best alignments. Even if I try to run it as a slurm job using separate nodes for each sample without -t, I still get only the best alignments. Both ways of parallelization failed to provide all alignments. Any help is much appreciated.

example code: bwa mem -a contigs.fasta -p 0_reads.fq -t 24 > 0_sample.sam

I notice that setting the -K option (bases to be processed in each batch) to a large value ensures reproducibility. However, will it solve the specified problem for the -a option as well?

Thank you.

Best,
Yazhini