Understanding ACES and ODT

In what context do you mean exactly? If you are creating a product for SDR displays your monitor and ODT should be set to that. I’m pretty sure that a display capable of 85% Rec2020 can accurately be hardware calibrated to it with Delta errors less than 2. You could set a display to native gamut but then you become reliant on the software managing the conversion in the case you want to view an sRGB image which you can’t do if it is not tagged or supported by the software.

Doing it on the display side of things removes that uncertainty because it can be calibrated and validated.

I think there are two main reasons. Encoding efficiency and grading predictability. Covered a little bit here. Difference between Virtual and Real Color Primaries - #3 by sdyer