|Unfortunately I ran into problems similar to yours at a refinery that was burning gas that emitted a different flame spectra and there was a lot of radiance at 3.9 micrometers. I could see through the flame but measured tube temperatures over 300 F higher than they really were. We discovered this by noting (in a vertical heater) that the furthest tubes ended up hotter than the ones near by.
The answer to your question is an expensive one. You need to get a spectroradiometric curve for that gas and if there is a wavelength that allows for it, get a custom made filter.
I have no experience with that type of heater so any industry experience out there - as in others who have done this successfully - should be of interest to all.
ANYONE ELSE WITH OTHER IDEAS?