Boulder::Blast is probably one standard way of parsing BLASTN results (and other tag/val reports), and will do a good job of simplifying the data held in the record, from where you could split the match and reference sequences into individual characters, and compare them, outputting mismatched characters (and their positions within the match).

You might also want to check out bioperl: parse BLAST HSPs which gives some example code on handling the records, to get at the info you want.

Hope this helps - have a go, and if you get into further trouble re-post and I am sure many a Monk will rush to your aid - remember to include input/code/outpur/error messages etc...!