Hi Ben,
I appreciate the effort you put into resolving this. Did you, by chance, just run the buffer again?
Did you have similar issues creating a buffer for the contaminated section of Emigration Creek? I don't recall so...
My guess is the buffer calculation just went awry.
At the scale of 2.5 km, the scale bar and measure tool should reflect the correct diameter of the buffer ring. I would imagine that the references you found to the geodetic distances simply means that the projection is taken into consideration when Arc does calculations and can effect the measurement. Curvature of the earth can be accounted for in all Arc calculations, something you can set up in the environment settings. But at the scale of 2.5 km, shouldn't impact your results.
Don't redo the lab, of course, but if you are curious, try rerunning the buffer tool again, verify your coordinate systems (display [data frame] et al) are in UTM and see what happens. I can maybe buy that there could be a pretty sizable discrepancy if the display data frame coordinate system were in the Web Mercator auxilliary sphere of the basemap??