Hi Nathaniel,
Been watching this thread. The only bits of info I haven't seen go across yet --
Antennas act very differently when the radiator part of the antenna is in close proximity to the RF ground. If you have modelled the antenna as being "in empty space", that will account for much of the loss seen when it is mounted to the sat, which in your design by definition is very close to the sat frame, and thus both the electrical and RF grounds. Have your modeller check to see if they have the sat frame, and any conductive panels, in their model; if not you'll have some (possibly large) matching adjustments to do.
A 1/4 wave piece of steel measuring tape needs to be at 90 degrees from all nearby metal to be an actual "monopole" antenna. If you lay it along the side of the sat, the nearby metal (grounds) will soak up much of the signal.
If I remember right, Prof Campbell's prior-to-CubeSat project (name is escaping me at the moment) incorporated a proven loop design. I don't remember the freq though. You might want to check your spec against that one.
If you haven't already, check the entire assembly (antenna on sat) with an SWR analyzer, or preferably a network analyzer, to make sure the transmitter is "seeing" a 50 ohm RF load at the input of the antenna line at your transmit freq. While this doesn't guarantee the antenna is actually an antenna and not a resistor, it does give a basic check that the matching network is close for your situation.
GL, Mike N2VR