The concentration varies considerably due to bees harvesting pollen from different flowers and from some batches of honey being mixed and some not.
Those are two separate worries connected by an "and". I'm not arguing worry #2 is overblown.
And it sounds like the problem of variance is that it isn't small. If one batch can be 5x as strong as another (made up number), then even a pretty good medicinal response curve leaves you vulnerable to that much variance.