Which is why a patent seems like a good idea, if your x-ray shotgun doesn't win a contract, but everyone starts using your x-ray scope, then you would still make a profit from your research. But if nobody wanted your x-ray scope, then you wouldn't get anything. This would give it a game-play reward for doing good research without just refunding all the research costs. As for what would be a valid patent fee, that would be a game-play issue. Make it too high and folk have little reason to ue established technology, make it too low and nobody will ever make a profit. Or you could let people set a price, so I could make a standardised accessory attachment and put a low patent on it, say 50C, I won't see any profit until it has been used in 21 designs by other people, but it is cheap an universal so that may well happen. I might like my extra-long rifled barrel, but it is a niche item that will likely see little use, so I could set the patent at, say, 750, so I get a healthy profit if just two designs use it and folk still get a small discount over researching it themselves, although they might just use a public domain barrel instead... Something like a single-shot light chamber might be good for a concealable weapon, and to get it to work effectively it may take a lot of time to refine the design, I might actually set the patent to be more expensive than the research costs because it is likely to see very little use and folk are likely to be willing to spend a little extra to save having to spend multiple turns refining their own version.
The obvious problem is that this requires extra tracking on the part of the G.M.. The custom version would require that every component have a patent rating. All variants would require that the researcher be tracked. Income would be coming from other players which would complicate things beyond just the G.M. handing out coinage. And if the players are not setting patent amounts then the G.M. has to figure out what would be good, probably basing it upon how many uses would be required to make a profit. And the current system works, there is little incentive to research, but if your design won't work without it then you are going to have to bite the proverbial bullet sooner or later...
An alternative approach would be that research gets you an income from elsewhere, maybe from the civilian market. Maybe something like <total cost of research>/(3/20) for 7 turns. You get a profit but it takes a while...
Patent system=complicated.
Income=takes the risk out.
No refund=whiners.