just ran the number again and today the direct and avg mult for best null / best high is 2.6
Then I realized ppl were talking about Ls, not NS.
avg mult is 2.17 better for LS than HS (again, for best offers)
direct mult is 1.37 better for LS than HS.
And yes volume is taken into account. I look at 1M LP point sold, normalized to the system (eg in NS it will be 3M25) , then the quantity of items you can produce with that number of LP and I multiply the price of the item by min(1, daily_sale/quantityproduced), which means if you can make 100 items with those LP but only 50 are sold daily, then the price is divided by half.
For the direct order, I sum up all the orders to buy that item, sorted decreasing, up to the X first, X being the quantity produced. I also do that for the items required (sum the lowest SO ).
Of course that means that if I require more LP then the prices decrease ; however I have a few bugs and that actually means that my program crashes 
I am comparing the LP worth, in term of isk, of the same mission.
When I say NS, again it’s the best NS agent, which changes all the time (sometimes guristas, sometimes sanshas), to the best HS agent , which also changes all the time.
I can’t guess what will be the prices tomorrow, so I run the number to the effective prices yesterday, and the direct prices today, so that if I run an agent now I know at least how many isk I can make per mission with direct orders, and if the market stabilizes (which I don’t expect) how many I can make tomorrow with the average.
Also this does not take geography into account, as it looks like the radius of the constellation is important, and maybe the number of jumps done in the individual systems (i.e. I got only two burners in Rens, when there was nobody almost in there). I use another program for this, limited to HS, that also takes into account the prob to go in LS.