why might scientists be wary of the results of a single experiment testing a particular treatment, even when the results are peer reviewed and statistically significant

The results may not be true for humans if the experiment used a model organism.

Explanation:

An animal model is an animal species used in research that mimics diverse human traits/features such as, for example, human response to certain diseases. Thus, animal models may provide useful information related to human developmental biology and human diseases (i.e., prevention, diagnosis, treatment, etc). However, in general, experiments involving animal models cannot be used to predict human outcomes. For example, in genetics, animal models exhibit different features to humans, including but not limited to genomes (i.e., different number and structure of chromosomes), genes, gene regulatory pathways, epigenetics gene regulatory mechanisms, etc. In consequence, animal tests may be unreliable and therefore their results cannot be directly extrapolated to human subjects.

the father in the first generation probably had the recessive colorblind trait in his x chromosome. if he and his wife both had the colorblind trait, even if she had 1 recessive and 1 dominant, then that trait would be passed on to their kids.

(since you didn't attach the family tree this is just an assumption)