Have you ever ran the test back to back and gotten the same exact readings? It doesn't quite work that way. I want to say even on 4G but I can say from personal experience any time I've run tests like this regardless of the type of network, I've seen big swings. I'd say any swing, big or small, is more related to an ever working network.
Your car comparison might have worked a little like a Faraday cage or antenna extension boosting or focusing the signal around you in some way and cutting back on interference. While 5G has already been deemed worse behind brick and mortar walls, metal may have a different effect. I'm just basing this on the evidence you've provided and not because I know anything. Guess the test would have been to run the test 3, 4 or more times in the car, looking at the variance, then comparing the same number of tests outside of the car. Then comparing the average that way.