I think it's very easy to get hung up on refuting or confirming the observer's explanation and ignore the observation. For as much as observations are worth, if someone says they saw a strange animal, I might believe them, if they said they saw Bigfoot, I wouldn't, but that's a very good moment to ask questions.
(Would you only take a cancer drug if the oncologist can explain exactly how it works?)
Explanations need to be tested to verify they are correct. Incorrect explanations delivered authoritatively lead to bad extrapolations, which in turn lead to incorrect procedures. So, if something has been observed to work but an explanation has not been verified, the proper response at the practitioner level is "This tends to work, but we aren't sure why".
Procedures are neither correct nor incorrect but rather effective or ineffective. Whether the explanation is correct or incorrect is irrelevant if the procedure is effective. Of course we would prefer to have both the effective procedure and the correct explanation but if we insist on only applying procedures with correct explanations we are not doing the best we can.
So, I would extend the proper response to, "This tends to work, we don't know why, but you should try it."