I'm skeptical of this method, so I did an experiment. I did three carefully controlled "ladders" from low charge weights to higher charge weights with equal step sizes. The goal was to eliminate sources of error, and presumably the velocity flat spot should be apparent in the data if it objectively exists. The first step in error reduction was to measure the resized brass at the shoulder keeping only one length plus or minus 0.001 inches. The goal of this is to keep the initial conditions of the case volume as uniform as possible. The next step is to weigh the measured brass and only keep one weight (less than 0.1 grains ES). Next it's important to use a very carefully measured charge, so I used a milligram scale and measured each charge to the kernel. I could distinguish between long kernels and short kernels in the H4350, and I used tweezers to select the appropriate kernel to get the correct charge. I measured the MV with a magnetospeed. The ladder was run three times at a slow rate (about 1 shot every 5 minutes), and I warmed the barrel with 10 shots of factory 6.5creed ammo at a similar rate.
results:
Each ladder exhibited what could be interpreted as a velocity flat spot between 2 or 3 of the data points, but it was not consistent between the three ladders. Additionally, when I averaged the three data sets together the result was a smooth line. The velocity increased monotonically as you add more powder. I think this method looks at random variation in the data and assumes that it's significant. I don't think the velocity flat spot exists. It seems to me that this method is like interpreting tea leaves.