The literature is genuinely mixed and the reason is selection, not measurement error. Dual agency transactions aren't randomly assigned. They cluster in situations that already differ from the base case: hot listings where the first buyer through the door writes, new construction where the builder's agent handles everything, and thin rural markets with few agents. Some studies find a small seller-favorable price effect, some find a discount, and a few find dual agency deals close faster at prices closer to list. Faster and closer to list is consistent with either story, which is why the effect sizes don't converge. Anything in the 1 to 2% range should be read as an average over a mixed population, not a coefficient you can apply to your 1.2m property.
What you can price is the process risk rather than the price effect. Your 15k inspection credit line is the better place to put the haircut, because that's where a dual agent's incentive bites hardest. A credit request has to be argued to the seller by the same person collecting the seller's fee, and a true dual agent generally can't advise you on how hard to push. Assume you're negotiating that credit yourself and budget your own inspector plus a contractor bid for anything structural, so you arrive with numbers rather than opinions.
The thing your framing leaves out is data availability. You're implicitly comparing your price to comparable sold prices, and about a dozen states are non-disclosure, meaning sale prices never become public record. In those markets your comp set comes from the MLS, and the agent controlling both sides of your deal is the same category of person who entered those comps. Ask what your source is before you rely on the spread.
Also confirm whether your state permits dual agency at all, since several restrict or prohibit it and the arrangement you're modeling may not be available.
