The feedpoint SWR will depend heavily on the type of feed system used.
The 4SqTL.EZ example model included with EZNEC, which uses one of the
all-transmission line feed systems described by "The Simplest Phased
Array Feed System - That Works", shows a 50-ohm SWR of 3.7 at the
feedpoint. Other feed systems are possible using this method, each of
which produces a different feedpoint impedance. The combination
transmission line - L network method described in Chapter 8 of the ARRL
Antenna book produces yet other feedpoint impedances, depending on the
transmission line Z0s which are chosen. There's no need to settle for a
feedpoint SWR in the thousands, and there's no need to alter your array
elements as a means of improving it.

If you use inverted-L elements for a four square array, it seems to me
you'd get a different elevation pattern in the various directions, and
that the azimuth pattern would be distorted by the directionality of the
elements. Have you modeled the full four element array, with the
currents phased as they would be for each of the four readily switchable
directions?

Roy Lewallen, W7EL

wrote:
Take a look at the far-field pattern and you'll see why I want it.
Also, it seems that inverted L's rather than verticals lower the
SWR from the thousands to something more sensible.

The Eternal Squire.