This appears to be a very poor estimate. My Garmin's VO2max estimate is more in-line (somewhat above*) my VDOT. For example, at ~20:15min 5k fitness, corresponding to a VDOT of 49.1, it will show my VO2max to be 53.
*We can expect the VO2max value to be higher than the VDOT, since the latter is a resultant 'aerobic power' metric from the combination of one's VO2max and 'running economy'.
However, I think Garmin's VO2max estimate (and 'race predictor') fall short somewhat, in that it estimates these values by (I think?) an average of the last five run activities. If those runs were performed over poor (slow terrain), in slow shoes, with excess clothing, etc., then it is going to arrive at a lower VO2max estimate (and race prediction), than if all of the five most recent runs were in optimal conditions (i.e., on a flat race track in good weather).
Out of curiosity, here is the first 1.5 mile VO2max estimator I found:
https://www.trainermetrics.com/fitness-assessment-calculations/vo2max-one-and-half-mile-run/As a comparison, it predicts one with 20min 5k fitness, possessing a VDOT of 49.8, with an equivalent 1.5mile time or 9'08", to have a VO2max of 56.4. It is probably looking at tables of population averages. Theoretically, Garmin (at least some of their fitness watches) may be doing something a little more sophisticated than that using FirstBeat's algorithms evaluating Heart Rate Variability.