Why Chinese Is So Damn Hard, Part 2

Alright, my last note attracted some dismissive comments from quite a few Chinese/Japanese-learning friends :) I feel like I need to be a bit more specific about a few keypoints.
(a) Language knowledge = character knowledge: Of course, I have lived long enough in China to realise this is not true for everyday communication with natives. As Xavier has pointed out, a relatively high level of communication can be achieved with just about 200 key sentences. My reasoning here is that I have personally reached some kind of plateau in my oral learning curve, and now I want to be able to understand more textual content. I want to be able to read blogs, tweets, forum posts and subtitles in Chinese. Thus the need to estimate what could be my goals and how realistic they are.
(b) It is possible to guess the meaning, and sometimes the pronunciation of a previously unknown character in a sentence: well, if you can do that, congratulations! I believe this character should belong to the corpus of known characters then, and that you should have a more accurate estimate of the number of characters that you truly know. Either you understand the meaning of a character in a given context, or you don't. How frequent are both cases? I believe the latter is way more likely. It would be interesting to think of a way to evaluate that.
(c) Twitter = biased source of textual content: Yes Thomas, in Twitter you need the context to be able to understand a given tweet (though that does not occur so frequently in my experience). That means the former curve is too optimistic and that you should drag it down to have a more accurate estimate of what you can read given the number of characters you know. This has nothing to do with the character frequency.
(d) Learning words with known characters is more efficient than learning new characters: Thomas, I am not sure the curve I have plotted can agree with that. What this curve says is that even though you might know all combinations for the 2000 most frequent characters, you still won't be able to read more than a tweet out of two.
I am surprised that people contest the implications of such an estimate, but not the experimental protocol itself (this behaviour is quite un-scientific). Nobody asked how I sampled the tweets or if the frequency table of Twitter characters is representative of textual content. There would be things to say here, but being more accurate here would again mean dragging the curve further down. This would not change the pessimistic conclusions I have already drawn.
Alright, I'll give you two good reasons for hope:
(1) The second derivative of the curve is positive from 0 to 2500 characters, where most students of Chinese are. This means the more you learn, the faster you learn.
(2) I am working on a revolutionary program to make your character learning more time- and effort-efficient :)