First off, I would want to know if the weighted centers of all the groups fell withing the other two groups for that same load. If not, I would add the other two centers to each group as phantom shots (hypothetical zero points) and re-evaluate.
That said I would be fairly certain something's jacked up with load B or the testing. A 9:1 ratio between largest and smallest group for a small number of 10 round groups suggests that the distribution of distance to aim point is modal rather than smooth. In other words, sometimes one thing happens, sometimes another rather than the usual small summation of errors that makes up rifle accuracy. Until that big issue was identified, I'd stick with load A for any practical purpose. There's less I don't understand and most things that need doing can be done with a 2 1/3 MOA rifle.