In old-school theory, the best results would be the traditional linear workflow and working in 32-bit floating point. This is the model of the traditional image processing programs like pixinsight, imagesplus, and partially in deep sky stacker. The idea in these programs is you feed the camera raw files to the program and the program decodes the raw data. You need to separately apply bias, field fields and dark frame corrections. What the typical workflow of these programs do not do is the color matrix correction required for Bayer Color sensors.
So if using the astro software, like pixinsight, you must do this by hand. Some programs, like deep sky stacker does not do this, so you would need to apply it by hand after stacking. Without doing the matrix correction, colors are bland because the filters in a Bayer color filter array have a broader spectral response than the human eye. Pixinsight "color calibration" does not fix this problem. As a result, colors from traditional processing are bland and people boost saturation to try and compensate. But because the color matrix correction is not done, the resulting colors are not accurate. As a result, the amateur astrophotography community has an inconsistent and bland color perception of the deep sky because the majority of those using the "traditional workflow" do not apply the matrix correction. But that isn't the only problem.
Telltale signs of poor processing resulting in color shifts are commonly seen in astrophotos, even by experienced astrophotographers. For example, as a nebula fades into the background, we commonly see interstellar dust go from bland tan to blue. Blue would be caused by very sub-micron dust particles and a particular geometry for scattering light from a star. In the common processing, especially by pixinsight users applying dynamic background extraction, the color shift with scene intensity can't be explained by known physics. And examining histograms the source of the problem becomes obvious: black point errors.
which again compares traditional liner workflow with photoshop ACR. Also see the section with Figure 9 which shows multiple people's attempt with linear processing (mostly pixinsight). Note the huge variation in derived colors with the supposedly "better" linear workflow. Again, no contest. 2b1af7f3a8