Two invented names, checked. The interesting part is what Synorigin did not do.
Whether the check would admit it could not find a close match, or strain to produce one so that it looked thorough.
Xantheon Galleria - closest matches
The gap
"Xantheon" is genuinely invented. "Galleria" is the derivative part, doing zero differentiating work.
Verdict
The name-word is original; the format is 90% a category default. Swap or drop "Galleria" and this closes clean.
Vrelkanth Sundergate - closest matches
The gap
"Vrelkanth" is invented and carries no baggage. "Sundergate" is doing all the work of sounding like a mall, built from the most generic naming formula in the industry.
Verdict
80% a naming-convention template - "Vrelkanth" is fresh, "Sundergate" is the part everyone else is already using.
Both results found something true at the pattern level - the naming convention - while stating plainly that the invented word itself had no match. Neither invented a fake competitor.
This behaviour is deliberate and is held to the same evidentiary standard as reporting a real match. A clean result is only worth anything if it is hard to earn: a tool that tells everyone their idea is original is not a check, it is a compliment.
It also shows a distinction worth naming. An idea can be part-original and part-generic at the same time, and saying so is more useful than a single number.