I suspect the UV sensor needs a high voltage because it produces a weak signal. It probably makes slight response to the gas flame which is dim blue. Suppose you test a variety of photosensors? Especially infra-red (although it may really be better at detecting a red or yellow flame since that is more easily visible.) Possibly combining two sensors of different wavelengths?
It's easier to detect at a close distance. If you need for the sensor to be this far away then consider mounting a lens so it concentrates light on the sensor. That is, if you can be sure the glass is the proper type to focus UV and/or IR.