There are a number of pix here:
http://www.sdsefi.com/air46.htm
The belly rad is very difficult to do nicely on RV3-9 due to the spar setup and lack of central tunnel like the -10. The -10 was no picnic either.
For the cowling and airframe you have, my advice is a single rad mounted under the oil pan fed by your shark mouth. If you can, build a dedicated duct which stays attached to the rad. This duct should diverge less than 7 degrees to the rad face if possible. If not, use guide vanes to turn the air without turbulence.
If possible, converge the exit duct aft of the rad and have a cockpit controllable cowl flap, this should reduce cooling drag measurably in cruise. This is harder to do well on an A model because of the gear junk in the way.
The rad shop of choice currently is
http://www.macsradiator.com in Portland. Not afraid to do aircraft rads and had done a number for other aircraft. When I get back from Reno, I'll try to work out a size and fin/ tube spacing. Be aware that without a good duct design, tightly sealed to your inlet, required rad size varies considerably.
My view on flat mounted rads is that they have high drag although some people are using this setup successfully these days, I have not seen any comparative speeds.
Oil cooling is critical on the Wankel so consider cooler choice and ducting carefully.