Let see if we can empirically improve the ACES workflow.
The EXR linear that you have are images that either are created from scratch (CGI) or they are camera capture file that has been converted. I'm curious to understand which ones they are.
I started with doing some internet research and I talked with Joseph Slomka (our Color Science guru and one of the ACES minds behind it) and this is the summaries of what we came out with:
Giving that ACES is a very wide triplet of XYZ values that can encompass all the color you can possibly see (and more outside of the boundaries ) and that the representation in the ACES EXR is in substance a photometric linear light, given that the image is from EXR in NUKE, is already photometric linear in substance as well, the main differences are:
-tonal range mapping
-RGB gamut mapping.
The former is tricky, the latter is more trivial: make a 3x3x3 matrice that map the RGB(Yxy) of your current color space (sRGB/rec709/P3DCI/ any) to the ACET triplet using the method described in the SMPTE RP177 (you can find the triplet here:
http://www.oscars.org/science-technology/council/projects/pdf/ACESOverview.pdf).
Ig you can give me some more info on your starting gamut (Yxy coordinates of RGB and W point) I can probably help you to build a 3x3x3 matrice
The tonal range is an evolving representation on the output of the image (I had a brain fart when he was explaining to me) HOWEVER for the practical purposes of your workflow, you CAN safetly assume that is a 1:1 representation.
so,
if you load the EXR as is in the resolve, setup teh ACES workflow and chose the display color space as rec709 (assuming that you have rec709 display) the density/tone of your signal should be ok, the color will be desaturated/whacked. (Unless Nuke 7 can map ACES directly)
Your next task is to make your IDT/ODT as simple as build a 65x65x65 LUT where the math inside is the 3x3x3 matrice above mentioned and it's inverse to map the gamut of your current color space to ACES and viceversa. This LUT is your implicit IDT and the second is your implicit ODT that you will have to apply as 3D lut in input (select in resolve the shots and apply input lut) and 3D lut in output (export with the ODT lut "baked in"). You can definitely build the LUTs in Nuke. The method is displayed in the SMPTE RP177 color space transformations: you need excell to build the math and you will have the 3x3x3 at the end...
\
Tell me where I lost you...