The numbers really depend on the terminology, and it is largely too inconsistent for my liking much of the time in a sociological label sense in so many of these studies (I'm not sure I've ever seen one explicitly detail their criteria). I think the rate of the umbrella term transgender is very high when applied by a second party descriptively, but that many people, notably men crossdressing, do not necessarily identify as transgender and thus the actual identification rate will be lower than the descriptive rate. This would include sexual fetish stuff that isn't key to identity typically. This would be the 5%+ easy number, and significantly higher still if you include the culturally accepted women wearing men's clothing(though for practicality sake, limiting that to more distinctively male dress, such as the suit and tie look, as obviously things like including t-shirts would drive that number up to virtually every woman ever).
As per the more common (perhaps layman's?) view of being transgender (transsexual, more habitual cross dressing such as with drag, and other very lifestyle/identity driving examples), I think the rate is probably more in the 1% range.
I think full on cross gender transsexual is probably half of that, going in line more with the .6% estimates. (And I believe many studies do probably start with the faulty transsexual = transgender premise.)
Also, I think that number is going to vary to extreme degrees based on location because of demographic clustering. In other words, walking around rural Indiana, 1 out of every 100 or 200 people is probably not transgender. But meanwhile, walking around Atlanta, you will get 1 out of every 50 people is probably transgender. (Though even then, the clustering will of course be more specific down to the neighborhood level, but just to keep it simple.)
