If your aim is to detect SNPs between these two genomes, then in that case you can denovo assemble one sample to generate scaffolds followed by mapping of reads of another sample on scaffolds of 1st sample. Using these approach you will come to know the difference between one strain against another in terms of SNPs.
But if you want to detect SNPs in both the samples with respect to reference genome in NCBI then in that case you can map the reads of sample individually on reference genome and can call SNP using their respective alignment (.BAM). Using these approach you will come to know the difference each strain is showing against NCBI's reference genome in terms of SNPs.
In any case, you can follow Variant calling for bacteria pipeline which I found worth in my study.