In-camera NR vs. separate darks

When I turn off 'long exposure noise reduction', and take some darks at the end of the session, the results are never as good as the in-camera noise reduction. Even considering that I could get twice as much light when I skip the in-camera NR (and therefore 1.4X better S/N, all other things being equal. it's still worse.

For quite a while I attributed this to 'operator error', and thought, with experience, it would be better.

But the noise is still worse using the classic workflow of separate darks.

So - another test.

Here are two exposures (each as part of a set). The camera is a Pentax K-1. The first is 15 seconds at ISO 1600 and f/10. With the in-camera NR, this took 30 sec. The RHS is the same exposure with in-camera NR turned off, and the 'master dark' subtracted (the master dark that DSS produced when I processed the full set of lights, darks, and flats)

I didn't just blindly subtract the dark from the light on the RHS - I adjusted the dark intensity to get the best improvement in S/N.

Then I used the app 'ImageJ' to get a plot of a line cut through these images:

LHS image:

RHS image:

I suppose I could draw ranges for the peak-to-peak (or RMS) noise and calculate a pair of S/N values, but by eyeball measurement, I'd say the LHS S/N is 30x the RHS value.

If the darks were as effective an in-camera NR, both images above should have the same S/N. (But the right produced in half the time.)

Here is a 100% crop from each image:

So has in-camera NR improved to the point that separate darks are no longer the best workflow?

Or is it still just me still not doing the classic worKflow correctly?

David