Yes, I answered my own question using information from here:
https://www.weather.gov/epz/wxcalc_altimetersetting (same as the link you gave)
That formula is apparently what MADIS is using and, as you say, using a fixed offset is only an estimate. The effective "offset" at my elevation (248 m, as I said, after carefully scanning the vicinity of my location on the map and adding 3 m because I'm on the 2nd floor - I gave MADIS 814 ft because they want your elevation in integer feet, so they have my elevation as 248.1 m) varies from about 28.9 mb to 29.9 mb over a roughly 40 mb range of station pressures. At the current pressure, it's closer to 28.8 mb whereas I'm using 29.5 (because of the 0.1 mb resolution), which accounts for the 0.7 mb discrepancy.
That said, maybe I should calibrate my equipment when the pressure at my site is "about average", say an altimeter setting of about 29.92 in Hg. That way I'll minimize the error in my reported altimeter setting relative to MADIS.
Or... instead of using the estimated offset from the keisan calculators, would it make sense to simply calculate an offset from an average station pressure (using the formula) and use that to calibrate my station? Or should I wait for an "average" day and then calibrate using the estimated offset?
Edit: actually, I think I'm already using an offset calculated at an "average" pressure, by using ISA conditions (temp and pressure) in the keisan calculator. Right now I'm reporting high relative to MADIS because the pressure is lower than "average" (ISA standard); when it's higher than "average", I should be reporting low.
But to set my ABS, I think I should wait for an "average" day, or else take the console to the airport and set it there using reported station pressure.
At least, I *think* that's right. If I have it wrong, please correct me.