My lender's version of projected ADU rent came out $780 lower than mine

The gap you hit has three separate mechanisms in it and they compound in a fixed order, so it's worth separating them.

The first is the source of the rent figure. Lender programs that allow projected ADU income generally rely on an appraiser's opinion of market rent on a standard form, not on the borrower's comp set. Appraisers pull from a wider geography and, on an unbuilt unit, are estimating subject to completion against plans and specs. A 15 to 25 percent haircut against an owner's own comps is common. Your $1,325 against $1,650 is inside that range.

The second is the applied factor. Programs typically discount gross market rent before it enters qualifying income to account for vacancy and operating cost. The size of that factor and whether it applies to ADU income specifically varies by program and has moved over time.

The third is the cap, which limits ADU rent to a share of total qualifying income rather than a dollar amount. Because it's a share, it binds harder on lower-income borrowers and on higher-rent units, which is the case you were in.

The order matters. Haircut, then factor, then cap. Running the cap against your gross comp number, which is what most people do mentally, overstates the result badly.

On confirming: get the loan officer to produce a written qualifying worksheet with a specific rent figure and the applied factors shown, and confirm the current program terms with your lender in writing, since these policies and their effective dates change. Neither the appraisal nor the underwriting factors are things you can negotiate after the fact.

On the contract side, a financing contingency written to a specific loan amount rather than to "obtaining financing" behaves differently in a case like yours, and how it's enforced varies by state. That's a question for a real estate attorney in your state.

The unconverted garage is doing more damage than you're crediting. An as-completed rent opinion on plans is the weakest input in the whole chain and it's the one everything else multiplies against. If the garage had already been converted and leased, the appraiser has an actual lease to look at and the number tends to hold. Different deal entirely.

I've been researching for six months and this thread just moved my ceiling number down on every deal I've looked at. @rena-expert the ordering point is the thing I'd have gotten wrong. I was applying the cap to gross.

Disagreeing slightly with the framing. You didn't fail at underwriting, you failed at contract structure. You built a purchase ceiling that depended on a number nobody had produced yet and then committed $8,500 to it. The financing math would have been the same in any market. What you controlled was whether to have hard money down before the appraisal came back, and a longer contingency or a lower earnest deposit was available to you.

I've bought small multifamily where the rent roll was already real and I still don't put serious earnest at risk before appraisal. The unit not existing yet makes that worse, not better.

@cairn is right that structure was the lever, but sellers with backups don't grant long contingencies and low earnest. Windlass was probably competing. The real choice was to bid a price that works on the appraiser's likely number and lose the house to someone who bid on their own number. Which is what happened anyway, just $11,400 later.