No, Microsoft, wrong conclusion. See, your data told you the $deity's own truth, that start menu usage has dropped. Most people pretty much use desktop shortcuts 90% of the time, so your stupid fisher-price jolly candylike tiles may look like crap but don't seriously impact that specific usage pattern. More accurate data collection won't change that.
What your data didn't tell you? That remaining 10% of the time doesn't just mean people "forgot" they had a shortcut and decided to use the start menu for the fun of it. Using the start menu drastically beats having to hunt down actual executables somewhere on the HDD, particularly for administrative-type tasks that might go six folders deep into the Windows directory, and have insanely long command-line arguments as a bonus (ie, a lot of the control panel apps).
Data doesn't equal knowledge. The stats can tell you "how often", but not "why".