I think it is because of the centering, rather than moving the balun further out. The balun on the stock design touches more than one point on the strips, so the contact point is not definite. The spacers make the contact point definite at the center between the two inner bays.In rabbit73's results it does show a freq shift [leveling effect] and a slight overall improvement, but is it due to the feed point centering or is it that the balun spacers moved the feed point further outward from the face of the wisker elements?
I retained the shape of the strips in order to be able to repeat measurements on the stock design if needed, and intend to (if I have time) substitue straight strips which I believe will work just as well.
The gain curve of a well designed 4-bay antenna rises slowly from the low end to the high end. This is because the elements and bay spacings become a greater percentage of a wavelength. My test results with the spacers seem to imply that. Without the spacers, there is an unexpected gain increase at the midchannels, which might be helpful to some viewers.
In any case, the modifications found so far improve the performance of the stock 4221HD, and make it much better than the original 4221.