Warning: This is long and might be completely uninteresting. It's also hard to make it make sense without visual aids, so it might be nonsensical.
tl;dr: I think my great-great-grandfather was either adopted or someone else altogether.
I got a new match on 23andMe not too long ago -- a 2nd to 4th cousin, the site said. Since the user name said TJCapello*, it became my closest actionable (i.e., non-anonymous and as yet unsolved) match on the site. I sent him the default "let's share DNA info and see how we're related" message, but -- as expected -- I didn't get an immediate response. His profile was new and contained no additional information.
I looked up the initials and refreshingly uncommon surname and, taking into account that he was male, I found his full name and location online with a quick Google search. I started drawing up a family tree for him based predominantly on his mother's obituary on Legacy.com (but also using pipl.com, Facebook, FamilySearch, and Ancestry), and I was delighted to learn three out of the four of his grandparents were Italian immigrants. I have only trace amounts of Southern European DNA myself and a tree filled with British and German names, so I focused my tree-building efforts on the non-Italian quarter of his ancestry.
Then I got another new DNA match, even closer this time -- a 2nd to 3rd cousin, it said. I quickly learned it was my previous match's sister (different surname, but Google knows all). Whatever my relationship to her is, it's the same one I share with him, so I figured I should be able to find our most recent common ancestors in the great-great-great-grandparent range or even closer (thank you, ISOGG).
I built out the English-sounding quarter of the Capellos' family tree until it should have intersected with my own. It even featured the surname Willis* like my own tree, and they lived in the Midwest, not far from another branch of my own family tree. But I couldn't find any overlap, despite my own Willis branch of the family tree tracing back to the 1600s.
I put this project aside for awhile, and I come back to it every so often. This wouldn't be an easy one to solve like I had thought. Either their family tree contains an error -- perhaps from an adoption or a non-paternity event -- or mine does. Or maybe that mysterious branch of my family tree that ought to lead back to New York where my great-great-grandfather was born really doesn't.
Recently I've started finding other DNA matches, on Ancestry this time -- all in Ancestry's "4th to 6th cousins" range, which tends to be a very loose estimate -- whose trees overlap with that same Willis branch that doesn't fit into my own. I've found upwards of five matches whose trees overlap in the same place, making them all second and third cousins of the Capellos, though Ancestry hasn't put it together into a "hint" for me yet because I sometimes have to draw up the family trees myself based on less detailed trees or user names alone. I appear to share about half as much DNA with those Ancestry matches as I do with the Capellos, which leads me to believe my family tree intersects with the Capellos' a generation more recently than it intersects with the others'. But that leaves me confused. Looking at their family tree, that means I'm descended from a Willis born in the early to mid-1800s. I already have all those slots in my family tree filled. I don't know how they could fit into my own tree.
That said, I don't believe any ancestor on my family tree is necessarily the right one until I have at least a couple separate (non-sibling) matches whose combined DNA and family trees support my data. The more distant the ancestor, the less possible s/he is to confirm. The more distant the cousin, the less possible s/he is to confirm. I'm in contact now with some cousins so distant that the relationship doesn't even show up in our DNA anymore, and I only feel confident of the relationship because of overlapping family trees and mutual DNA matches within those same family trees.
Logicking It Out
Here's the deal with the Willis branch of the tree in question: It shows up in several reasonably close DNA matches' trees, so I assume it is how I'm related to them. It's possible I'm wrong, but it's unlikely. In order to fit it into my own tree however, something currently in my tree must be wrong. First, I know the Willises are connected to my maternal side because my paternal uncle on Ancestry shares zero of those matches with me. I also have enough known DNA matches at this point to draw the conclusion that several specific ancestors on my tree must be accurate. I can verify my mother is my mother, I can verify her parents are my grandparents, and I can verify my great-grandparents too. I have enough reasonably close DNA matches backing up my data that I feel confident about six of my eight maternal great-great-grandparents. I even have an Ancestry "hint" that aligns another more distant cousin with ancestors of one of the two remaining great-great-grandparents (I feel less certain because it's only one match and a distant one at that). That would leave Jack, my great-great-grandfather who supposedly came from New York.
Jack is the brick wall of the mystery branch of my family tree. I have no DNA matches to support him, and many hours of research have yielded no indication of who his parents were, which makes it exceptionally hard to find DNA matches that would support him. His wife, my great-great-grandmother Emily, was from rural Illinois, within a 45-minute drive of the Willises. According to census records, she was twenty years younger than Jack and had their first child -- my great-grandmother -- when she was 28. They'd supposedly married two years earlier, but I have not been able to find a marriage record, though I found one for her first marriage easily enough. Lots of my ancestors crossed state lines to marry though, so I'm not even sure where to focus my search. Could Jack have been my great-great-grandfather but actually been adopted? I would think this more likely if he didn't claim to have grown up in New York, over a thousand miles from the family to which I'm trying to connect him. I could be wrong, but I don't think adoptees were moved that far from their birth families in the 1850s. Could my great-grandmother have been a non-paternity event (NPE), meaning Emily was impregnated by someone who wasn't Jack? If that is the case, I'm still not sure who my great-great-grandfather would be. There isn't one specific "most likely suspect" in the Willis family tree, either based on DNA or based on relative age and geographic proximity.
My closest DNA match on Ancestry whose tree contains the Willis line has several matches in common with me. A few of them also contain the Willis line, but several don't have detailed trees, nor are they related to the entire cluster of other Willis descendants, though they are related to each other. My next step is to build family trees for the ones who don't have them yet, or whose trees only have a couple of names, which is most of them. My hypothesis is that the ones who aren't mutual DNA matches with the Willis cousins will be related via an adjacent family line -- perhaps the Thompsons. Thompson was the maiden name of my closest Willis cousin's great-grandmother. If I'm right and they're connected via an adjacent family line, it would tell me which generation connects me to that family tree -- the generation containing both the Willises and the Thompsons (or whichever adjacent family surname) rather than an earlier generation.
In case you're wondering why I would put so much effort into something that matters so little, please understand THIS IS MY FAVORITE KIND OF PUZZLE. I have been waiting for something like this to happen ever since I solved the "who is my biological father?" puzzle, which was at most a 4-star difficulty on Dell Logic Puzzles' 5-star scale. I find few things as gratifying as solving logic-based puzzles, and solving this one will create an even bigger hint toward solving other genealogical puzzles, of which there are two more I've been working on for months. I've written about Aida, but there is another one I haven't even mentioned yet (she self-identifies as Cherokee, but her DNA is 99% European), and the solution to this Willis puzzle will help me towards solving both of them via deductive reasoning. In short, I'm doing this for fun.
*Not his actual name.