How are you checking yeast health? From what I've read, the methylene blue stain is not accurate below 95% viability.
I've read that too. It is "not accurate" to some level. Even the American Society of Brewing Chemists calls it not accurate.
If by "not accurate" you mean that it is not exact then I agree. There is some error, but that error can be determined so that the accuracy is known.
It is very dependent on the strain and how much Methythele Blue that is used. WLP004 and WLP566 stain just fine with 0.03% MB, but EC-1118, and WLP650 are better with 0.06% and 0.1% respectively. It's also very dependent on the operator. MB staining is very difficult to use on some strains, and there doesn't seem to be one concentration that works for every strain. If you use too much it will actually start to kill some of the weaker yeast cells. The method isn't perfect, but it has it's place.
I have a blog post on this coming up.