If you are automating the generation of the graphs, making assignment of which teams are par 1 the first step of that process is not too onerous. In terms of accuracy, you would probably have to use a ppg table rather than the actual table to reduce the effects of the variable number of games played.
Of course if you were being grandiose, you could use an updating full table - using the matches from the current season, filling in the gaps with the results from the previous season - using relegation/promotion team substitution where appropriate. That would factor in the current form, and still use a full season worth of results.
Mind you, I could not be arsed to set that up, and I my "being arsed" threshold is quite awesome. The current approach is simple, yet good enough.