Hi jking, and sorry to take so long getting back to this. I think that the reciprocals were probably the problem, and that your reacting to them perhaps confirmed to Google that you were aware of a desired effect... and that when the links appeared to backfire, you tried to undo the situation, not once but twice, which made it look even more manipulative.
On the original links, I believe it's not just the crosslinking alone, but also the rate of acquisition. Read martinibuster's last post in this thread....
What is Unnatural About Certain Links? Google Penalizing Unnatural Links Mar 20, 2012 http://www.webmasterworld.com/link_development/4431410.htm [webmasterworld.com]
In the thread, martinibuster links to a thread in our Supporters area (subscription required) where, based on a comment Matt Cutts posted, a bunch of us discussed a crosslinking situation that may have trigged a false positive for link selling...
New Alert? "Webmaster Tools notice of detected unnatural links." Google Webmaster Alert http://www.webmasterworld.com/opengoogle/4249049.htm [webmasterworld.com]
20% of your product pages, IMO, linked all at once, could well have triggered some kind of drop. That's a lot of what is called "co-ordinated action", enough that it's very clearly not natural.
10 days before the drop was probably enough time for Google to have processed the links and to react... and in fact if Google was doing what I suspect they
might have been doing, that would have been appropriately prompt reaction time. This is conjecture... but I think that Google may have wanted you to notice cause and effect, and created a strong negative response to see what you then would do. See this thread, which discusses that possible Google strategy. We've never had absolute confirmation that the patent was in use, but I believe I've seen the effect happen enough that I think this is what was going on....
Google's Rank Modifying Patent for Spam Detection Aug 18, 2012 http://www.webmasterworld.com/google/4486158.htm [webmasterworld.com]
Your nofollowing and then removing the links could have signalled to Google that your actions were countermeasures to the negative Google reaction to the links.
In this case, I'm thinking that your using nofollow crosslinks to start with would have been the only way to have sent the proper signals... and that leaving the links as nofollows after your rankings didn't come back may have been your second best course (as it might at least have suggested that you were linking for users). Again, this is conjecture.
I'm not sure what I'd do now, except that I wouldn't fiddle any further with the links. You didn't mention, btw, whether you'd gotten a note in Webmaster Tools, so I'm not sure whether there's any penalty for which to request reconsideration.
If Google is using the rank modifying patent, though, it's unlikey that they'd send messages telling you that it was in use.
Conceivably, a Penguin algo refresh might solve it. John Mueller of Google has recently suggested that they might reward Penguin-affected sites which show signs of improvement before the refresh is run. I'd take this as a strong hint to continue improving user experience and content on both sites.
I'd also be very cautious, though, about things like getting links from common sources for both sites. I'd keep them very separate.
