Clearly, among the 146 data points in this meta analysis, many are non sensical. Maybe the author could have spent some time weeding out the worse results.It can be seen that VO2max decrease is never bigger than air pressure decrease (for your info: air pressure at 1000 m is about 900 hPa (10% decrease vs sea level), at 3000 m about 700 hPa (30% decrease), at 5500 m about 500 hPa (50% decrease). It's basically proportional for lowest dots - almost linear dependency between air pressure and VO2max meaning that oxygen amount being the bottleneck in human engine.
BTW: Holly, molly! Who had a 15% VO2max decrease at circa 5200 m ASL? Superman Lopez, Superman from Krypton or some extremaly unfit fatty?