Theoretically, Gartner might have used data from retailers (how many N900/Blackberry/iPhone they sold) and use this proportion, along with known figures for some devices from manufacturers. I have an N900 which came from country A (from Nokia retailer here?) through country B (middle-stop) to country C (the end user; the device brand new). The path took more than a year (warranty expired). It would be interesting to know how Gartner and Nokia counted it, to what year and country they attributed it. Though it cannot explain all the discrepancies in data, of course...