Away for a day and look at all that happens! Amanda, if you really want self-expression, you should hear me sing. I'm thinking of trying out for The Voice. I mean, when I sing all the dogs in the neighborhood--like up to seven houses away--start barking and going crazy. Neighbors have pleaded with me to stop. They think it's annoying the dogs, but I think they love it! My best performance so far was during a Caribbean cruise a few years ago. Late evening, big crowd, really stormy night at sea driving everyone indoors, karaoke in the main bar. I sang The Wreck of the Edmund Fitzgerald. The audience was positively stunned into silence!
Ahem. Okay; back to business. I hope nothing I wrote earlier made you think that what you had was not a Wiki-useable 3rd cousin "confirmation." Reinforcing again that I can't speak to the WikiTree policy, but you have a sourced paper-trail match to Carolyn, the cM amount falls within a 14% probability as shown at DNA Painter, and so I think that means you should be good to go. And Kay's idea about Greg's confirmation tool is a good one. But someone else needs to speak for the WikiTree DNA Project about policy/guidelines. It's Wednesday, and I'm only ever wrong on Wednesdays. Or was it Fridays? I forget...
Shortly you'll have a private message with an unsolicited and probably useless "care package." Hope you're fine with a ZIP compressed file and that you use Excel. If not, don't bother downloading it. :-)
There are a few things I routinely do when examining small(ish) segments. GEDmatch is kind enough to tell us the origin (if they know) of the raw data file, and in this case yours is from a MyHeritage v2 test and Carolyn's is from an Ancestry v2 test. Not all AncestryDNA v2 tests are created equal, but the best I can do is assume the "template" of the initial, announced v2 release.
Given that, the net message is that the comparison itself is weak in terms of examining the same SNPs, and that the segment may be a bit longer than advertised. Can't say for sure, but it's a maybe. Taking the Build 37 start and end points of that shared segment at GEDmatch, of the 10,818,299 base pairs encompassed MyHeritage tested 2,280 of them (about 0.021%) and Ancestry tested 2,570 of them (about 0.024%). That's roughly in keeping with the overall percentage of SNPs our microarray tests look at compared to the entire genome.
However, the MyHeritage and AncestryDNA tests examined only 813 of the same SNPs within that segment range. So now we're down to 0.0075% of the segment size. Not much coverage. The GEDmatch free one-to-one analysis tells us that it used 796 SNPs for the comparison, so it excluded a handful of the SNPs in the original raw data (every kit uploaded there goes through a process they call "slimming," where they streamline the size of the data used for final comparisons by removing some of the SNPs they deem to be least useful for the purpose).
I also took a look at 20 SNPs flanking either end of that segment. There were breaks in both cases, but based on what SNPs the MyHeritage v2 test checks, five of the 20 in front of the segment and three of the 20 at the end were also tested at Ancestry. Flipping, six of the first 20 at Ancestry were also tested at MyHeritage, as were four of the last 20 flanking SNPs.
That doesn't inform the comparison that GEDmatch reports; it is what it is. But GEDmatch ended up comparing only 1 base pair every 13,591. Makes me want to get back on my whole genome sequencing soapbox.
Another...well, not anomaly, exactly, but very possibly an illustration of that disparity of SNPs tested by each of the companies, is that I broadened the search to a start locus of 50 million and end of 70 million on Chr 1, and then ran a matching segment search at GEDmatch for both you and Carolyn. You came up with only five "hits," Carolyn with 35. Only one of them did the two of you share in common. He's in the spreadsheet and, all things considered, might be worth contacting.
So... You have a valid match. You have a researched paper trail. You have evidence that the level of the match does not rule out 3rd cousins. I believe (but never trust me on WT policy matters), that you should be able to proceed to "confirmed with DNA" status indicators based on that. And I'll push a few chips to the center of the table and bet that if we had better coverage of that region of Chromosome 1, you might end up actually sharing more DNA with Carolyn than 15.3cM.