A few things you can try-
1) First I'd try keeping both omnis in positive polarity, and do some stereo widening on the omni pair before mixing with the center subcard. The Subcardioid center mic is going be be providing plenty of monophonic mid signal, so by changing the balance of the omni contribution so that it is providing mostly Side info, the combination of the three mics should be much more open and well balanced. Otherwise you end up with too much Mid buildup, sounding overly closed and narrow, as baffled omnis are relatively Mid-heavy through much of the spectrum to begin with and mixing in a dedicated center mic doubles up on the Mid content.
Widening is a Mid/Side type manipulation, typically done to a Left/Right stereo signal. Depending on what editor you use, it might be available right on the channel strip as a simple knob twist (in Samplitude you can do it that way at the object level or at the stereo-track level in the mixer). If not, here's a good free VST plugin for Mid/Side adjustment-
http://www.voxengo.com/product/msed/. It can also be used to convert from Mid/Side to Left/Right or vice-versa like most Mid/Side matrix plugins, yet this one has two M/S matrices available in series, so you can input Left/Right stereo and it will convert it to Mid/Side, allow you to readjust the balance ratio between Mid and Side, then covert back to Left/Right at it's output. Using that, try increasing the wideness (more Side, less Mid) of the omni pair. Both the input to the plugin and the output from it will be Left/Right stereo, and the Mid/Side adjustment will made entirely within the plugin. Make the omnis as wide as it will allow- that is all Side and no Mid from the omnis, and mix that to taste with your center subcard. This is doing a Mid/Side mix without having to do any mixer routing stuff, and such a mix should limit phase-cancellation issues since all the Mid is coming from the center mic, and all the Side from the omnis.
Extra credit- You can also try widening the omnis to a somewhat lesser degree, in which case you'll be mixing some Mid contribution from the omnis with some from the center mic. You'll probably want to use less center mic level that way, and it's possible you might get some phasing, which might be good or bad sounding. But try it with the omnis "fully widened" first and compare against that.
2) You can do the routing manually. Which is actually the same thing as above, but done a different way. We're turning the omni pair into a dipole, at least partially, in place of a figure-8 bidirectional Side mic. This is similar to what you are doing by inverting one omni channel or the other, but does so in a more balanced way by inverting one omni channel, summing them, mixing that sum with the center channel to produce stereo-Left, and mixing the polarity inverse of that sum with the center channel to produce stereo-Right.
To do that, first mix the Left-omni signal and the
polarity inverted right-omni signal together. Listen when you do that and carefully adjust the level of one channel or the other until you find the greatest degree of cancellation of bass frequencies in the resulting sum. The greatest cancellation of bass frequencies indicates the most optimized and balanced di-pole sum. Save the resulting mono mix of left-omni and polarity-inverted right-omni. You now have two saved mono channels to work with- the center mic (the Mid-signal), and the sum of the omnis with one channel inverted (the Side-signal).
Now open your mixer screen with three channel strips. Put copies of your omni-sum mono-mix on two of those channels, and your center sub-cardioid mic on the third. Pan the center mic channel to center. Pan one of he omni-sum channels fully Left. Pan the other omni-sub channel fully right
and invert it's polarity. Then bring up all three faders. You now have control over Mid verses Side balance (center mic fader level verses that of the other two channels in combination) and over left/right balance (the difference between of the two omni channels, without changing the center level).
Notice that in doing this you are inverting polarity twice- once in creating a mono-dipole Side signal from the two omnis. and once again in the Mid/Side to Left/Right matrix. Other than the part about "balancing the level of the two omnis to find maximum cancellation", that's the significant difference between these techniques and what you have already tried by inverting polarity of one omni channel and then mixing the three original microphone channels directly. This way you'll gain similar openness, but without the bass being reinforced on one side and cancelled out on the other, and with potentially improved imaging.
Both of the above techniques will only be effective at producing and preserving stereo information above the frequency at which the Jecklin disk begins working. The addition of the center mic facing forward potentially provides more pickup of direct sound with increased presence and clarity (good) yet doubles up on the monophonic Mid information which decreases stereo width over that provided by the omnis alone (bad). These techniques essentially reduce the Mid contribution from the omnis to make room for the center mic contribution. To further increase the sense of openness and low frequency stereo-ness, you might try cutting the bass of the center subcard while boosting the bass of the omnis, in a complementary way so that the overall bass level remains the same. That will increase the difference signal at low frequencies, similar to using a larger baffle, which should increase the overall sense of openness.