I feel that the right approach (at least it worked for me) is to tell the matematician - "see, this term is very close to zero for all practical inputs we can see in real life (proof of which will be left as an exercise for you), so - as you can see from the structure of the equation - we can safely drop it without introducing any significant error to the end result. Yes, the result will be a little bit less accurate; no, no one will notice. And if you happen to encounter a high-energy (or whatever) problem, you can always go back and use the full equation". Explicitly accepting that each handwave introduces errors and imprecision but that sometimes those errors are below what we can measure (or care about) should be acceptable enough for a math-minded person.