Out of 300 ways extracted from address database I see about 40 entries (counted by hand, too lazy to build a proper query) being very alike. Luckily, most names decline and "street" and, say, "boulevard" have different grammatical gender, so it's harder to make a mistake.
Why yes, Mr Vanbrugh[1] did live in the castle on the corner...
address says "Sokolska 20". Okay, let's guess from the other metadata that the city is actually "Prague" and not "Ostrava". No such house number at this address, perhaps should have been "sokolOVska 20, Prague"? Wonderful, that exists, solv- oh wait, do we have two different houses with the same house number? We do, multiple clicks apart, and both are legally correct.
In other words, that's a very good example you've used, exactly because it shows the underlying assumptions as faulty.