Looking at the maps for Japan on OSM and iOS 6, if they are using OSM data it must be a very small subset and they're adding a ton of crap over it.

For example: they usually have one label for a train/subway station which is well placed and then a second label for the same station 200m to the south-east. OSM has nothing like this. They must be cobbling many datasets together with no fact-checking...

To what degree are they using it? Not a large one.