The 14% figure lacks context to be truly useful for outsider interpretation.
Consider this problem as an analogy: take a book and randomly open two pages. Your task is to compare these two pages and express their similarity as a single percentage. There are many ways you could go about it: you could just brute-force it, comparing letter-by-letter, or you could compare individual words, sentences, or attempt to gauge the meaning of the whole page at once. You could give a higher similarity score for words that have the same letters in a different order, or you could prioritise synonyms.
The different approaches are all valid in their own right and can be practically applied depending on your goal (e.g. hunting typos or checking accuracy of a translation), but the single number you end up with won’t really be useful unless you also say which method you used.
The same applies to genetic information. There are many ways to compare genetic samples, the results of many of which can be expressed in percentages, and they are chosen depending on the task (comparing species from different kingdoms of life, comparing closely-related species, comparing different populations of the same species, determining ancestry, identifying mutations in different cells of the same individual, etc).
There certainly are reliable genetic markers that can be used to determine ancestry (it wouldn’t be labelled “black” or “white”, but something along the lines of “possibly Senegal or generally the West African region” or “North European, most likely Sweden”) and I’m sure the test in this story did find relatively recent partial ancestry from Sub-Saharan Africa, but without more details I wouldn’t dare to speculate which or how many of the guy’s ancestors were African.
