Comment 1 for bug 1549487

So this is actually a lot more difficult then you're probably assuming. The data we're using for graphing most things is being processed in a few ways. The most obvious of which is that it's being resampled (at least once) into a different resoltion based on the selector. We're also doing some mild interpolation for missing data points. The numeric data we end up graphing doesn't really map back to a single run.

What we're really looking for here though is to map the metadata for the runs we're graphing to the data in the graph. Assuming you're talking about the per test graphs I think if we added an optional filter selection to graphs to select which results we're showing that would accomplish the same goal. It would require a new rest call for each filter (and for a while we probably could only support a single filter) but that would at least give users an indication of where the data is coming from.