Interesting, but I'm not sure it's meaningful. Consider the following.
1) Their sample data is based on your product usage and people who use rescue time are probably heavy cloud users. So population set is unfairly waited toward the cloud.
2) Microsoft's get's the majority of their users and revenue from enterprise level sales. Google customer on the other hand is primarily the consumer and small business. This is important because in most enterprise level environments users can't install an application like rescue time.
3) Lastly people double dip - at work use outlook for work and gmail for personal.
It would interesting to just look at productivity apps like word processing or presentation with referrer values from .com addresses only (not logins) . The trending might be very different.
I also figured that their data is biased, but the interesting part is that the trend line goes counter to that bias. Say that RescueTime is heavily used by early adopters that are comfortable with cloud-based services, the same people that are likely to use Google Apps. In the beginning, you'd expect that to bias the results toward Google. But as RescueTime broadens toward mainstream appeal - which is already happening, I was an early adopter but then dropped it when it got too corporate - you'd expect Google's market share to drop, as RescueTime's share expanded to include more Microsoft users. Instead, you see the opposite, where Google's share is growing even though the population of RescueTime users is expanding into domains typically held by Microsoft.
Our users may include more tech savvy early adopters than the general population, but capturing that audience is how tech markets are won. This makes our data more relevant as a leading indicator, not less. If you are a major software vendor, and you lose the early adopters, you should be worried.
I voted you up because you made an interesting point here about early adopters, but it would have been nice if you made it in the blog post rather than in the comments here. It would make you more credible. Along a similar vein, you might consider labeling the y-axis on your graphs with something like "% of RescueTime users" instead of "% of computer users." My message is that if you were slightly more qualified in the post, you wouldn't have 75% of the comments telling you that your post is misguided, and instead the overall discussion in the comments here would be much more intelligent.