Agreed. And that's because Garmin's VO2max is a HRvPace look-up table. It simply uses your %HR and a "known" value for different associated %HRs to determine the relationship to pace. For instance it says marathon %VO2max is around 82%. Based on the relationship between HR and %VO2max, wherever your recent runs have you at a 82% VO2max is going to spit out a piece of data for the Garmin to draw a line. Then, it'll use those paces at those set %VO2max values for 5k, 10k, LT, HM, M, LR, Easy type paces to determine a VO2max estimation. This is similar to what I do for myself:
This is real data. I tracked the relationship of my HR vs Pace over time during a Spring season. When one is able to run at a faster pace at a set HR, then the ability to race faster goes up. In lock-step when I saw increases in my ability to race faster, the Garmin VO2max increased. When the end of the spring season came, my HRvsPace relationship started to reverse direction. This was because under those new hotter temperature conditions I was no longer able to race as fast. The Garmin VO2max declined. So under the conditions I was in, I was less capable of running the same speed then when it was cooler. The Garmin VO2max reflected that. I wasn't losing fitness per se, but I was not capable of running the same speeds in the conditions under which I was now running.
In laboratory conditions, when the temperature of the room is manipulated, then it appears the oxygen consumption ability of the runner can be manipulated as well per the research I found.
So the Garmin VO2max is not measuring true VO2max as a measure of oxygen consumption. But, when temperature changes both the Garmin VO2max (HRvsPace relationship) and true VO2max (laboratory tested oxygen consumption) go up and down in a similar (but not necessarily the same) manner.
I'll use this chart as an example. The average runner's body weight was 72kg in the study. So a 3.2 L/min oxygen consumption at 4C at 45 min was a VO2max of 44.6.
Absolute VO2max (in Liters) = weight in kg X VDOT/100
But at 31C at 45 min the oxygen consumption was 2.6 L/min. For the same body weight, that equates to a VO2max of about 36.1.
So in this laboratory setting, when altering the temperature from 4C (39 F) to 31C (88F) the VO2max of the same runners was altered from 44.6 down to 36.1 (or a drop of about 8 points in a matter of a week or so dependent on the testing temperature). And this drop (8 points, or 19%) in VO2max simply couldn't be accounted for with detraining since Jack Daniels data suggests it takes about 72 days of inactivity with no leg aerobic exercise to reduce by that much. So yes, it would appear a significant drop in true VO2max (oxygen consumption) is possible in a very short period of time when accounting for changes in ambient temperature per the research I was able to find. Now that doesn't necessarily mean Garmin VO2max measurements drop at the same rate or in the same way as a true laboratory VO2max measurement, but it does appear that what you've experienced in your Garmin VO2max value is replicated in this particular laboratory setting for true VO2max oxygen consumption.
Now the reason this came up was a discussion about a measure of tracking fitness gains/losses over time. If I were to compare the 2017 summer to the 2018 summer's VO2max number, then we could get a pretty good general idea as to whether I've made training gains/losses. If I were to compare my beginning of Spring 2017 to end of Spring 2017 under similar weather conditions, then I could get a pretty good general idea as to whether gains/losses have been made. But I think we can both agree, that the data becomes harder to interpret when you compare different temperatures (or across different seasons/the whole year). But it is a reflective measurement of what you are capable of in that very moment in those specific conditions in which you just ran.