That's wrong. Breakdancer takes alignment as input (BAM). Correspondence between your sample(s) and reference is present in that alignment.

You can run breakdancer on multiple samples by having a configuration file that gives information about each sample you want to analyse. The easiest way to create such a file is using the bam2cfg perl script

bam2cfg.pl SampleA.bam SampleB.bam SampleC.bam

You should see separate lines in the config files giving the stats for the three bam files

Then when you run breakdancer, each SV it detects will be annotated with how many reads from the different libraries have supporting evidence for the SV. You could then parse out SVs that are only supported by one sample.