You're aligning multiple query sequences to one reference. A single measure of similarity for multiple queries does not make sense. So the answer is: you don't want to do that, using BLAST or anything else.

Maybe the best would be to generate a density plot or histogram of the e-value distribution from all the assembled contigs? BLAST does not have the functionality you describe.

You could use BLAST -m8 output and tabulate the bit scores - that might be a somewhat meaningful metric but it wouldn't really stand up to scrutiny

perhaps MAUVE would give you some kinda single metric: http://gel.ahabs.wisc.edu/mauve/

First thing is there is nothing called a stupid question.

The tabular outfmt 7 of blast combined with pident (check the stand-alone manual), gives best results but do use the bit scores for comparison.

If you would like to use the results for further data merging then I guggest you to use XML output (outfmt 5) and then parse whatever you need.