Beware of ethnicity estimates, especially the new timeline one from 23 and Me. (Here’s another genetic genealogist’s case study of how unreliable 23 and Me’s new Ancestry Timeline is.) It’s nearly impossible to identify from the DNA alone the date one of your ancestors was last “purely” from any one country or ethnic group. There are two issues in play.
First, our genes are wildly mixed up. While different ethnic origins can sometimes be assigned to different DNA segments, this only tells you about the background population from which those genes derive. It cannot on its own identify when a particular person came from that country. The article linked above gives the example that someone with 39% British Isles ancestry might have a parent from England, or, as is actually the case, a bunch of really distant British ancestors up several different lines. If genes from those people happen to be inherited in sequence on a number of different chromosomes, it might appear that the genes originated as a single unit in the recent past rather than the reality that a bunch of different segments were inherited from many different people who ultimately came from the same population centuries ago (i.e., roots in New England).
The second issue is historical. National borders that exist today did not exist a few centuries ago. Migration has always taken place. Putting these two together, you can understand how someone of French extract—let’s call her Angelique—might have a mix of genes that appear to the test to be Scandinavian and Mediterranean (Italian) rather than Western European. For Angelique, the current French border obscures the deeper history of Roman occupation and the Viking settlement in Normandy. Even if everyone in Angelique’s document-able family tree spoke French and lived in northern France, Angelique’s genes suggest that she happened to inherit more genes derived from Roman and Viking fore-bearers than, say, Germanic (Frankish) ones.
Taking this example one step further, it’s impossible for commercial DNA tests like 23andMe or AncestryDNA to determine on their own (without documentary evidence of some kind) whether Angelique’s Mediterranean ancestry in fact came from an underlying Roman genetic pool or instead originated in a smaller population descended from, say, a troop of Italian craftspeople who intermarried with the local French population after migrating to France to help build a 14th-century castle. The history is simply too complex and the DNA too fragmented to tell the difference without serious scientific study of particular genes or without much larger pools of historical DNA to compare with.
In the end, population history and population genetics make something like 23andMe’s ancestry timeline an impossible endeavor. It may make good marketing, but it doesn’t make valid history.