Unexpected result of Gamut Mapping in RCM/DWG

I have found a puzzling behavior of the Gamut Mapping FX when used in a Resolve Color Managed Project in DaVinci Wide Gamut.

I have for years used the Gamut Mapping FX in Saturation Compression mode to tame chroma overshoots. It does a nice job. I had not been using any Resolve Color Management.

With Resolve 17.x, I have been experimenting with the new Resolve Color Management and DaVinci Wide Gamut. My input and output color spaces are Rec 709 Gamma 2.4.

If I add the Gamut Mapping FX to my node tree, it immediately clips all values over 100 nits. This is not what I expected.

Is this expected behavior, and why? Is there any way to get it to work in RCM+DWG as it did before, just compressing chroma and not clipping luminance? It works as expected in RCM in Rec 709 Gamma 2.4.

Test system:

Resolve Studio 17.1.1 Build 9
Desktop Video 10.11.4
Decklink Mini Monitor

Win 10 Pro 1909 build 18363.1379
ASUS X99 Deluxe II
Xeon E5-2696 v4 2.2 GHz
64GB RAM
2 x 1080Ti
Nvidia studio driver version 461.92

If I add the Gamut Mapping FX to my node tree, it immediately clips all values over 100 nits. This is not what I expected.

What happens if you bring up the luminance value in the Gamut Mapping panel? Can you restore the lum back to where it needs to go?

What happens if you bring up the luminance value in the Gamut Mapping panel? Can you restore the lum back to where it needs to go?

Hi Marc. Thanks for the suggestion. I found that by iteratively adjusting the max input luminance and max output luminance, I could eliminate the clipping and bring the luminance level back to where it should be. However, the saturation remained about the same. It was necessary to also reduce the saturation max to get the FX to do what I wanted. All in all, it was a very fiddly exercise compared to the automagic operation of the Gamut Mapping FX in Rec 709.

It seems to me that the whole idea of RCM and DWG is to handle all these transforms behind the scenes, so I don't think this FX works properly in this scenario.

I need to reread the order of operations and re-watch Alexis van Hurkman's Ripple training on RCM to see if I'm missing something obvious.