The study compares Democratic turnout relative to Republican turnout in the 2004 Presidential election to Democratic turnout relative to Republican turnout in the 2008 primaries.  When you compare that data, only 3 states showed reduced Democratic turnout: Arizona(which is fairly close to the parity line), Michigan, and Florida.

So, even though Democratic turnout in the illegitimate early contest in Florida may have been record-setting, it's also significantly out of line with the pattern that's been observed nationally.  An estimate made in the study comes up with the idea that, based on these national trends, turnout in Florida for the Democratic primary should have been 2.85 million voters, instead of 1.7 million, a difference of 1.15 million voters.  

For Michigan, it finds that turnout should have been 1.305 million voters, instead of 590,000.  The Michigan number is remarkable, because, as the study concludes that Michigan should have had a turnout of 1.3 million and only got 590,000, there is a difference of 715,000 voters.  In other words, based on the study, more would-be Democratic voters stayed home than went to the polls.

In total, that's an estimated total of 1.865 million people who would have voted in the Democratic primary, had Florida and Michigan held contests that counted.  Perhaps these states can somehow come together and find a plan to run a real and fair primary contest before June 10th, and if so, we might be able to discover just how close these numbers are to reality.  But, if the states cannot get their act together and do just that, any argument to seat delegates based on those illegitimate primary contests must also consider the near 2 million people who didn't vote not because they didn't want to, but because the contests were determined before the fact to be meaningless.

If you'd like to argue about the data, that's fine, but I think the data shows either:

1. Florida and Michigan's turnout was depressed due to the knowledge that their primaries would not count
   
2. Florida and Michigan have grown significantly more Republican since 2004, despite Michigan having Granholm win re-election by a significant margin in 2006 and the election of a few Democratic members of the House in Florida in the same year

Susan MacManus, a political scientist at the University of South Florida in Tampa, largely agreed. She said the controversy already had dampened enthusiasm among Democrats, who turned out in record numbers in January.

Florida Democrats, she said, "are very frustrated because this was a year that they really felt that the sun was shining on them. They would not only win Florida for their presidential candidate but pick up some [state] legislative seats and congressional seats. They were just on cloud nine and now, all of a sudden, it's like a storm cloud is over Florida."

Your recitation of the "1.7 million" talking point is unpersuasive to anyone who gives serious consideration to the issue, because, of course, an absolute turnout number has absolutely no bearing on the proposition that some potential voters did not vote based on their reasonable reliance on representations that votes for the presidential candidate would not be used as a basis for the seating of delegates.

I see you have finally made a responsive argument, which is that the disenfranchising of the 1.7 million outweighs "bending of the rules".  My response is that the "bending of the rules" actually amounts to "breaking of the rules" and one of the rules you would like to break is the rule that in a democratic election all stakeholders should know what the votes are for at the time of the election, and that should not be changed after the outcome is determined.

Taking your argument from the realm of abstract principle to the precise matter at hand, a more precise summation of your argument is that disenfranchising the aforementioned non-voters is justified by the enfranchisement of the 1.7 million, and that argument is self-refuting.