It is reasonable to assume that Nikon build their scanner to align the light sources/filters with Status-M curves which are representative of absorbance spectra of color dyes.
I doubt that. When Nikon built their scanners, they were limited to whatever spectral sensitivities the semiconductor industry (which in part they are/were themselves, btw) could muster. This was far more limited than what's possible today - a lot has changed in two decades.
Anyway, if you take this argument about curve matching and think it through, you at some point realize that the next digitization method will just create another interpretation of the same color negative. And there's no telling which one is the more 'correct' one. If you get down to the gritty details, you end up realizing that you're balancing a number of technical artefacts that mostly stem from physical and chemical limitations of materials used, and there's just no way to figure out which of these limitations contribute or degrade the end result.
Look at the typical overlap in blue and green curves (magenta & yellow dyes), for instance - that's not something that was originally intended by film manufacturers and they've worked hard to reduce this crosstalk as much as possible, but depending on how you scan or print the negatives, some of the effect will still be there. To an extent, you could say it's "intended" - after all, the film was designed and accepted for/by the market with this flaw built into it, but
exactly how much of it is "supposed" to carry over into the end result? And on which end of the curve (the blue/UV bit vs. the green side)? So you may decide to take color RA4 printing as the gold standard, on the assumption that CN negs and prints formed a single imaging system. Only to realize that CN paper also has its own idiosyncrasies that carry over into the end result - and to what extent is that desirable?
To further complicate matters, whatever optimization you may figure out for a particular kind of film may not work optimally for a different dye set.
The main takeaway is that each method has its own pros and cons, and you end up having to compromise. You take one seemingly obvious starting point (in your case, the status M curves vs. a random set of dye density curves) in the hope/expectation that you can figure out a good solution based on that. But once you dive into it, the whole exercise explodes in complexity.
One can make this really, really complicated and spend years in the books and setting up lab experiments to figure out what approach works 'best'. Or one could be pragmatic, run a couple of tests with equipment that's close to hand, and see if the results are acceptable. If so, settle with that and move on.