Natural non linear systems are self correcting and a single high standard deviation event is more likely to followed by an opposing event of similar magnitude in the opposite direction. (mean reversion)
Besides, each individual sample they've taken represents an average over a 1200 year period which introduces more variability than the single study addresses. It's like me trying to guess your weight and saying it's somewhere between 50 pounds and 5,000 pounds. I'm almost certainly right, but it doesn't tell you a thing.
Personally I don't think it's such a great leap suggest that those high levels are due to industrial output. That seems to make sense to me because 5STD's is REALLY high.
But getting from there to anywhere is meaningless based solely upon this data. It's going to take a lot more than that to get where they want to go.
(data set for reference: http://cdiac.ornl.gov/ftp/trends/co2/vostok.icecore.co2) I think you missed the point. You need to look at the sampling methods (finely slicing pieces of ice and measuring the CO2 in the slice) and determine the error bars for the reading by considering the length of time that the measured sample represents. I think you will find that a sample that represents some distribution (possible Gaussian) of several centuries of actual CO2 measurements will have an error bar considerably larger than the 5 STDs you attributed to the current reading. Also bear in mind that the old the ice core, the more compression and the longer the interval that will be included into a single reading.
Unfortunately for us, the solution boils down to the argument I am having with cogitator which is: can the CO2 spike or not? If it can, how much and how comparable to today's spike which we can all pretty much agree has some anthropogenic component to it. If natural spikes occur, how much can be hidden in the poor sampling resolution.