My poorly-founded speculation: Girls may be more likely to try to solve problems with help from fellow humans. This help is unlikely to be available in a controlled study. (It's also frowned upon or unavailable in a number of academic or professional contexts.) Controlled studies are, however, likely to have humans present, blocking the bits that step up and say "okay, I'm doing this by myself now".

That's a curious hypothesis. The corollary of the "males are incapable of asking for directions" stereotype could be the stereotype "females are incapable of navigating without asking for directions".