ok two things . . .
first, i found an error in my equation. instead of:
Torque * rps > = Mass * Acceleration * Velocity * (1 + efficiency) / (2 * pi)
it should read:
Torque * rps > = Mass * Acceleration * Velocity * (2 - efficiency) / (2 * pi)
sorry!!! ill correct that as soon as i can . . .
ok so your RMF with the correction should be 7.44 ft lb / sec (the RMF for one motor should be half that, since its differential drive)
referring to this datasheet:http://www.robotmarketplace.com/images/mabuchi550.pdf
now looking at the motor . . . so for torque, you dont want to calculate for stall torque, what you want is torque between maximum efficiency (58mNm at 21840rpm) and maximum power (407mNm at 12000rpm), depending on your application. If power or motor overheating is a problem, you want maximum efficiency. If torque or motor cost is an issue, you want maximum power.
So the RMF for the motor would be either 15.57 lb ft / sec at maximum efficiency or 59.99 lb ft / sec at maximum power. It looks like you can run at maximum efficiency with no problem. What happened was the speed you listed is for zero torque, and the torque you listed is for zero speed (not possible
Now for your other question about gear reduction . . .
Gearing is when you change the torque to speed ratio, so the RMF is unaffected. If your output will have half the rpm, but twice the torque, RMF doesnt change. However, since you have added gears, you have decreased the efficiency (reducing effective RMF). In my experience, homemade gear setups are low efficiency, so its better to buy a motor with a built in gear head.
in case you havnt seen this, here is where I give efficiency estimates for different types of gearing:http://www.societyofrobots.com/mechanics_gears.shtml
hope that helps!