I'm very happy to beta something in this area. I could go on about it for ages! though am not sure what else to usefully say at this point.
I sympathize with the problem of interpreting latency. Bufferbloat is distinctly detectable, painful, and common enough it should be part of any speed test. Sample the latency a few times before and during a download, check for an excessive difference. 100ms is the maximum the buffer "should" be set to; beyond that it starts to be a problem. The pain means it's a good reason to consider changing connection... The most simple, reliable fix is to use a higher speed connection. OpenWRT is an awesome fix if you want to set up a router all of your own... the fun starts with choosing a router that's supported and yet still for sale.
I guess other issues with latency or loss are hard for a convenient web test. I would have enjoyed seeing idle latency during all my "auto" tests. Whether it would have tell me anything useful is another matter.
A 100ms spike on the connection is a slight surprise and not entirely welcome . I assume this wasn't during a download test. I know downloads caused an average around 100ms, which is understandable.