False positive and false negative rates can be different. Trusting that when it says bad, it means bad can be consistent with doubting when it says good
If a device can tell you something is maybe poisonous, a positive result is going to lead to very different behaviour than a negative one, even if they’re equally likely to be accurate. This is especially true when your priors (ie research with accurate devices and knowledge of combustion) lead you to expect (prior to measurement) that it will indeed be poisonous.