Sounds like a case of "bad vibrations". The bullet is exiting the barrel at a time of maximum whip. The whip timing depends on the longitudinal shock waves travelling down the barrel, being reflected back to the chamber, reflected again etc etc. And such a "resonance" load is very sensitive to small changes.

You need a load that avoids the resonance points. If you are in the valley between two resonance peaks, then the POI will only change very slightly with load variations, but if you are around one of the resonance points, then the POI can vary quite sharply. It is difficult to find the best load, but easier to find a couple of bad loads and then pick a value in between. Run a series with varying loads from your fire-forming load up throught your bad load to the specified maximum safe load. The best load is that range where the group position varies least from one load step to another. That may be difficult to tell, but plotting a graph of group size and position against load will show it up. For instance, with my M1917, with 46 or 49 gn of N140 the grouping is about 3 times larger than anything in the range 47-48 gn. Same method ought to work with your Hornet (much smaller loads, of course!)

Patrick