ASHRAE 90.1 2010 versus ASHRAE 90.1 2016

You are correct. US DOE does a determination of the savings for each update to 90.1 relative to the previous. I have not seen a determination report that skips one but you might be able to infer an average savings by extrapolating from two determination reports. 90.1-2016 showed an 8.3% energy cost savings compared to 90.1-2013 and 90.1-2013 showed an 8.7% cost savings compared to 90.1-2010. So the "blanket" value of 17% will only hold true in very limited cases. The actual savings difference will vary all over the map from below 0% to over 30% depending upon climate zone, building type, and numerous other factors related to particular EEMs. You can find these reports here - https://www.energycodes.gov/previous-determinations

Be very careful in swapping to ASHRAE 90.1-2016 the changes in the App. G methodology result in much greater differences in savings than what is quantified in the determination studies.  Without doing the math to see the potential increase based on renewables, I doubt you would get more points using 90.1-2016.  Here is one study looking at differences in savings between ECB and App. G in 90.1-2016.

https://technologyportal.ashrae.org/Papers/PaperDetail/10855

Dave, that's a useful anecdote. We've been expecting things like this for v4.1 in EAc2 and EAc5 where you can apply renewables in multiple spots. 

Back to the earlier discussion on code determinations, PNNL does those based on their prototype libraries of models each updated to the newer code and against the prior one. These models are prescriptively compliant with the base code. Then they apply factors to account for each building prototypes' likelihood of being constructed in the new code in each region to forecast the new code's overall impact (See Table ES-3 for building type weighting https://www.energycodes.gov/sites/default/files/2021-07/Standard_90.1-20...). When you're looking at modeled under ECB, you get to fuel map like-for-like, which is potentially different than the PNNL Determinations. And now switching to 90.1-2016 Appendix G PCI/BPFs you get fuel mapping based on climate zones. You will find yourself in WI comparing all-electric buildings to a nat gas baseline per Appendix G and using costs and GHGs. So again don't expect % savings (or % more stringent) from the prototypes to directly apply to your project. 

The TAG is anxious to hear more war stories, so do share them.

If you're still curious, you should check in with Mike Rosenberg for any ability to test your initial theories on creating a % difference from '04 to '16. You can still find most of those prototypes online--even for the old models. I think it's likley of limited benefit, but I won't stop you from trying:

You could just pull out medium office buildings in CZ 6A, so you're not seeing a weighted avg from other CZs or building types:
https://www.energy.gov/eere/buildings/new-construction-commercial-refere...
https://www.energy.gov/eere/buildings/commercial-reference-buildings
https://www.energycodes.gov/prototype-building-models