Server comparisons and WPT slower than the real world

Given that the speed/CPU of the server is important, how can accurate comparisons between locations be performed? Some servers are older and slower than others…

Also, the load times for a page are slower than a typical user would experience in the real world. A typical page a normal laptop on standard wifi connection i find to be 2-3 times faster than web page reports depending on which server is chosen.

Is a fudge multiple the only way to solve this? for example, dividing the metrics by a certain number?

I had Cable selected which is a throttle. Choosing native connection removes the throttle. Although the results are better, they are still double what i would expect.