Or, to paraphrase a professional contributor to a genealogical DNA textbook, "Admixture (ethnicity estimates) is the most eye-catching aspect of autosomal DNA testing. It is the feature companies highlight the most when advertising their products. For that reason many people test solely for the admixture results but, in reality, it is far less useful for genealogy than DNA match lists."
And several pages later:
"Admixture estimates vary between companies because they each use different reference populations. They are accurate to the continental level, but cannot be taken too literally at the country and regional levels.
Be especially cautious with small percentages as these can be more likely to be erroneous." (My emphasis).
I read a paper on the ISOGG wiki some time ago, which showed that the "reference populations" being used by some testing companies in relation to certain regions can be as little as 20 people or fewer!
Small percentages "attached" to certain regions may not be indicative of any recent relatives originating from the region at all. It can simply mean that you have a trace indication matching a group of people who are also thought to originate from the same area, or who are most likely to be found in that area.
I've also read that approximately half of the population of England have some Scottish, Irish or Welsh "ethnicity" in their admixture according to Ancestry, but it doesn't mean that their ancestors ever necessarily lived in those countries. A proportion of the people who settled in England also migrated here from the same regions as those who settled in Scotland, Ireland and Wales, and whose admixture can today be found in a majority of those populations, but the truth is just as likely to be that the Scottish, Irish or Welsh admixture is an indication that all of our ancestors from several hundred to 1000 years or further back in time were originally from the same region in continental Europe, and which e.g. Ancestry now ascribe to Scotland, Ireland and Wales as that is where the majority of that admixture in their current reference populations relates to.
Ancestry ascribe 22% and 16% of my admixture to Wales and Scotland respectively. LivingDNA, who are generally regarded as having the largest UK reference population (and therefore the most accurate admixture "guestimations" for the UK) ascribe 11.9% of my admixture to the Welsh borders, and nothing at all to Scotland or Ireland. They give me 3.9% northwest Germanic, which includes the southern Nordic region and is fairly consistent with my maternal haplogroup, which indicates ancient migration from that area into northern Germany and then to France and eventually the UK, albeit that is also supposition. Scientifically informed, but supposition nonetheless! The rest of my LivingDNA guestimate is English through and through!
From what I know of my ancestry through paper based research supported by DNA matches, I have absolutely no connections to Scotland, although my maternal grandmother's family originated from the Welsh borders. Having fair hair, blue eyes and pale skin that burns in front of a light bulb I have always thought of my ancestry as Nordic, probably Viking. That might be the case in the mists of time a few thousand years ago, but my maternal and paternal family names are both suggested to originate from the Normandy area in France, which does rather fit with the migration routes proposed for my maternal haplogroup (Russia->Scandinavia->Germany->France->England).
But then I could probably find theories to support any admixture estimate I was given by different testing companies if I looked hard enough. I prefer to use individual DNA matches to inform or support conclusions drawn from documented research. Admixture or ethnicity estimates are best taken for what they are - estimates - and mostly with a pinch of salt, since unless you have individual matches to MRCA's from those regions, they aren't really of much use for evidence based genealogical research anyway