They usually run batches on existing DNA samples to update it so their results once their databases are updated with more accuracy because some ethnic groups share a lot of DNA whereas others like Finns, Ashkenazim, Sami, Japanese, etc. have very distinct genetics that make it easily distinguishable because those ethnic groups have very little admixture in comparison to a lot of Arab, European, or Asian ethnic groups who have had thousands of years of mixing with different tribes and ethnic groups. My Unassigned DNA has shrunk to 0.1% as they were able to assign the rest to Broadly Middle Eastern then to Persian then to Uzbek.
We have over 20,000 genes so even if you inherit only a small percentage of genes from a Norwegian ancestor then you still have chance that that "ghost" gene might reappear because 0.75% is a significant amount of genes that are European that can affect phenotype and overall health. During apartheid South Africa, there were several cases of Afrikaners with black ancestry from 1700s being born with mixed race skin and african hair so their status as white was stripped and were given mixed race instead. I do agree with your point about the inaccuracy of the smaller numbers because I'm 0.1% American Indian and were talking maybe 20 or so genes that indicate that which could just be a fluke and it's actually just Northeast Asian or something similar lol.