I was mapping paired ends from Illumina, with option -i PE 250,15, using Novoalign V2.07.13. That is, the average length of the library is 250 bp, with a standard deviation of 15 bp. I understand that novoalign looks for "proper" alignments up to 6 standard deviations away from the mean template length, in both directions. In this case, that would be between 160 and 340 bp. However, I see that some pairs are mapped much closer than expected (more than 6 SD below the mean) and still they are considered "proper", according to the flag set in the SAM format.
Is this behaviour expected? If so, why? Would not it make more sense to call those pairs inproper?