Despite the last few days when the pool worked well I think the pool PPLNS stats, the pool luck, is wrongly calculated. As I said before I think the pool luck between September 24 and October 24 was arround 80%.
I think you have overlooked just exactly what the "Graph" is charting. It represents a comparison between the estimated per share earnings of a pool which would have paid PPS minus 2% fees or the estimated average of actual per share earnings of the BTCGuild (With PPLNS, each open shift is rewarded the following:(Block Value + Transaction Fees) / 10) - Pool Fee (2%)).
As it would break down in a per (3?) shift time points for the graph and numerically calculated per shift % as determined by the position of (24H / 3D / 1W / 2W / 1M / 3M / All Time).
So what's wrongly "compared" is your idea of the graph and what the graph actually tells you.
If you would like to figure out strickly bare Pool Luck, you can get the latest 200 block numbers and time stamp of when they were found under the "pool stats tab" sub-table "Block History". If that doesn't go far enough back for you; there are plenty of other websites that provide a full listing of the block history and who/when blocks paid too, when the receiver's name is known, as BTCGuild's has been, for a long time. Then you can divide the dates and time however you like from and to whenever.
PPLNS involves variance: If the pool is lucky, you will earn more than normal. If it's unlucky, you will earn less.