What Garmin says is perfectly consistent with what I said. Garmin should understand this. They have been in the business a long time. The first paragraph is irrelevant here. The receiver computes position in WGS 84 cartesian coordinates, transforms those to the ellipsoid and, finally, if desired to the geoid or at least calculates and reports the geoid - eliipsoid difference which is based on location based interpolation between carefully surveyed benchmark points. The reason it is irrelevant is because the first altitude reading (at startup) will be subtracted from all subsequent readings. As an example I read the following altitudes from my iPhone (indoors - it's cold out there today - but next to a window), at 1 minute intervals 270,273, 273, 271, 272, 271, 270, 267, 268, 267, 267, 267, 273, 276, 277, 279. The actual altitude of the phone was 290.8 ' so all those numbers are in error. Were I the SOLO I'd subtract the first reading from each and obtain AGL altitude estimates of 0, 3, 3, 1, 2, 1, 0, -3, 2, -3, -3, -3, 3, 6, 7 and 9 feet. Those really aren't too bad and certainly good enough to assist in checking on the barometer or even use in a pinch if the barometer failed in flight. If I were the systems engineer that's what I'd do. Note that this is exactly what is done with the barometric data. The bias from the fact that SOLO is not at sea level on a standard day are removed by subtracting the altitude measured at initialization from each subsequent altitude reading.
The reason that this works is that the constellation does not change much over the course of a 15 minute solo flight so GDOP doesn't either and, unless a satellite is rising or setting, the estimated height AGL isn't going to either. Note further that SOLO can obtain GDOP, or it's component, VDOP from the GPS module. It can probably also obtain the covariance matrix of the satellite psuedo range residuals or at least the residuals vector from which it can determine the uncertainty in the altitude measurement and thus decide what weight to give the GPS altitude data. This is not to say that SOLO does or does not use GPS altitude as I have no clue as to how the thing is engineered. But I do know that engineers really hate to throw away data.
My iPhone said that the uncertainty in altitude was from 10 - 13' over that. I don't know whether that is one sigma or 2 or whatever. The standard deviation in those altitude data is 3.8 ft.