Here is my VVE table, this is after months and 100+ logs of refinement. I have been using Bluecat's tool to calculate the table. First I use the VVE editor in HP Tuners to copy and paste the scanner feedback. Then I use the VVE editor to interpolate horizontally (since I can more safely assume linearized behavior between RPM, rather than interpolate vertically since I think you get more non-linear behavior through RPM). This get it more filled out. Then I use Bluecat tool to calculate the coefficients and if any come back above the integer hard limit in the tune file, I have to smooth. I try not to smooth too much because I think it takes out what you are trying to do. What I like about the Bluecat tool is that it will adjust the zone boundaries as well as a variable to get even more accuracy:
Attachment 99232
It may be worth it for me to go back to the stock VVE and do exactly as you said increase the 40-50Kpa area a certain percentage, decrease the 50-60kpa by a certain percentage, then increase the 70-90KPa a certain percentage. When you watch the O2 or fuel trim feedback in the histogram and watch the cells its referencing, you can see it kind of go in a circle or that J-hook think like you said. The drastically different cells it references, along with all the other modifiers very well could lead to some negative feedback loop causing instability which causes the timing to spike around.
Would you mind sharing what your VVE table looks like?