"It's just "flattening" any camera response curve the .raw file have had"
Yep
When the image is flat, you color correct the image with the "ColorChecker Camera Calibration" app, and then fix the values with Photoshop/Lightroom.
You end up with the real life albedo value that you can use as pbr reflectance in the high pass filter technique.
There can still be pixels that clip black or white in the picture, but that's because they are in shadow or overexposed. We only care about the proper exposed real life value of a material, we need cross polarization to get the "real" real life value ;)
It is known that every chart differ, and as time goes by you get shit and scratches on the swatches. I wouldn't trust my life on it if your client is strict about the color, but it is good enough to get values for albedos. Maybe you should try SilverFast, I have heard good stuff about that scanner / calibration software.
Photographers use the chart differently, they use it for white balance and to get the correct colors with "ColorChecker Camera Calibration". They don't need the real life values, and the 100% calibrated images look pretty washed out.
If I remember correctly I had some problems with Corona and the "ColorChecker Camera Calibration" app. Because Photosop/Lightroom is looking for the camera model in the EXIF Data. And there is no Canon/Nikon etc data in the Corona image. I tried to fake it with a EXIF Data editor but failed. I haven't tried again after this old test.
I would put my money on 32bit linear render, fix the values with CameraRAW and then tone map with ArionFX/Nuke/Fusion.
If I remember correctly this test was done with tone mapping, and I lost a lot of detail on the bright curtains and in the shadows.
Now that I think about it, what a weird test, so unorthodox. Is this technique really better then a 50% grey sphere in the digital world ? Maybe I should do a retest when I get time and include a 50% sphere!