Here is a histogram of characteristic accelerations from the SPARC data (our data fit by arxiv.org/abs/1806.06803). For each galaxy, g+=x*Vf^4/(G*M). The range of the x-axis is set to match their Fig. 1. No fitting of any sort has been done. The scale a0 is in the data. 
Vf is the flat rotation velocity. M is the baryonic mass - the sum of stars & gas assuming fixed M/L for all galaxies. G is Newton’s constant. X=0.8 accounts for the geometry of disk galaxies (they’re not spherical cows and we can tell. See Binney & Tremaine). There at no knobs.
Lacking knobs and making the simplest possible assumption about M/L, the result just falls out. This has been shown before, eg arxiv.org/abs/1102.3913. If you add a lot of knobs and get a different result, all it means is that you’ve added too many knobs.
They grey histogram is all the data for which we can make this plot. The blue histogram applies mild quality controls: distance uncertainties < 20% & no galaxies with the lowest quality (Q=3) flag in the SPARC database. The more accurate the data, the more clear the result.
The width of the histogram is dominated by measurement uncertainties and the inevitable scatter in stellar mass-to-light ratios.These suffice to explain practically all the width.Hence the only interesting conversation being whether there is a tiny intrinsic width, or none at all
