J depends on wa and wa is proportional with L.
Bigger L means bigger wa and smaller J.
You can lower J (changing from "total uplift- anchorage required" to reasonable "partial shell uplift") by increasing wa, which means an increased L.
L can be increased by increasing annular thickness, because the calculated L is between two plastic hinges; a thicker annular means bigger L between plastic hinges. However, this procedure has limits: 0.035D is a limit of L, shell thickness is a limit of bottom thickness.

As limit, L= 0.035D is equivalent to wa=201.1HDGe (in SI) or wa=1.28HDGe (in USC Units). See API 650 (E.6.2.1.1-1a)/ (E.6.2.1.1-1b). In fact it is a limit of the quantity of liquid that has credit to resist overturning.

See also the basic document you can download from
http://mycommittees.api.org/standards/techinterp/refequip/Shared Documents/seismic.pdf; just study it!

The limit L= 0.035D has been established by Robert Wozniak in "Lateral Seismic Loads on Flat Bottomed Tanks" Water Tower, Nov 1971. I haven't that article, I cannot help you more.

In fact, why are you so concentrated on this limit? Are you ready to accept a model with double plastic hinges in bottom (on uplifted shell part) but not a limit of this model?

Happy New Year!