Citing personal experience is completely irrelevant to this kind of study. Firstly, the people you know do not make up a statistically significant quantity. But more importantly, 'people you know' is an incredibly biased sample - people tend to associate with others of similar socio-economic class and education level, so if you can find the US on a map, it's almost certain that the people you associate with can also find the US on a map. The problem is that you're excluding the population segments that you don't associate with - you aren't factoring in those who are homeless, mentally ill, those who never attended class or effectively skipped high school for whatever reason. Once you add those in, the figure of 1/3 might not seem so crazy.
However, I'm actually familiar with that study, and there's a context to it. People in the US were asked to identify the US on a GLOBE, immediately after being shown a standard flat map that is commonly used in the US. Any flat map is forced to greatly exagerrate the size of some regions while shrinking other regions, in order to represent a globe on a 2D page. In the US, flat maps almost always adopt a perspective that puts the US in the 'exagerrated size' section, showing the US as being much larger, proportionally, than it really is (they have to show SOME area as larger than it really is, so they choose the US).
What the study showed is that when people switched to using the globe instead, they were thrown off by the fact that the US is considerably smaller than their flat maps indicated. 1/3 couldn't do the conversion. Yes, it's still rather stupid, and worrying that they couldn't just recognise the shape of the continents for North and South America. But it isn't as though they didn't know where the US was in the world - the problem was adjusting from the flat map to the globe.
