I think simply deleting that first erroneous Tilt reading fixed the issue. As soon as I deleted it, the Attenuation reading was correct. And it wasn't immediate, but the ABV reading eventually corrected after deleting that data point, probably within a minute or two. That seems to fall in line...