I'm kinda surprised you didn't know this already. Some tests over-estimate, some servers just can't handle fast speeds, so you do indeed have to decide which to trust.
For years I couldn't get a good result on Thinkbroadband, speedtest.net is very hit and miss based on the server and time of day, fast.com I used to think was reasonably accurate but its not reporting great so far on Gigabit, BTW is known to have been bad for years.
In fact, Thinkbroadband specifically advise to use their download files to manually test if you're unsure of the result. I also have Librespeed on my VPS and also run iperf3 every once in a while from there into my network. Even iperf3 for some reason will occasionally give weird results just on the LAN.
One thing with the ping time is some tests monitor the latency during the test (the useful way to do it), but some do it when the line isn't loaded (not so useful). Plus the latency will vary depending on if the test actually bottlenecked on your connection, or somewhere inbetween.
Overall its why I use a variety of tests in my
broadband history diary. Those results are never from one test alone, I will do several, maybe over a few days, and cherry pick the one I think represents the most realistic result.