This may be simple or impossible depending on the graphing framework. The plots as they are currently are being generated are not very helpful. It plots the "totals" from each of the 6 columns but it does them all on the same axis which is maxed out at 100. In most cases (including the one on the "install" site) there are more than 100 lines of code which means that line isn't even plotted. One way to attack this problem is to have a number of plots that each have a few of the results with common "units".

I would say the following share "units":
Ratios: LOC/Method, Methods/Class
EntityCounts: Classes, Methods (debatable, classe count will usually be far outstripped by method count so it would probably be preferable to put them on sepearte scales)
LocBased: Lines, LOC, Code LOC, Test LOC

As I've thought about it I think the Ratio group may be pointless anyways because the ratios are all presented in integers which would make the graphs very unappealing (mostly straight lines with large jumps). The EntityCount group is more useful but mainly to help someone new to the project get a feel for the overall project status, during initial development or adding of major features you would see these plots trend upwards but during a maintenance phase you would see much flatter lines (since most changes are bug fixes).

I think the most valuable would be the Lines group. More specifically I think that the most useful single plot would be:

Code LOC (lines of production code, shouldn't increase much during testing)
Test LOC (lines of test code, the ratio is obvious since they are on same plot)
Non Code Lines = Total Lines - (Code LOC + Test LOC) (this is an indication of how well commented it is)

In summary I would say that overall, plotting all of the data returned by the stats task is actually probably less useful than just plotting the "Lines" plot I outlined above, having multiple plots or axis is probably more effort than it is worth.