The distortion appears in the top and bottom edges of the trapezoid. With no distortion the edges would be straight lines like this: 
The last pattern picture you posted looks a lot better but it still shows some distortion.
What is happening with the trapezoidal pattern is a plot of the audio input vs. the audio output with the output in the form of the RF envelope. Think of plotting the input vs. output of an audio amplifier. You would see a straight line tilted upward or downward. Here, since the envelope has both positive and negative going excursions with the audio we get both negative and positive sloping straight lines (if not distorted) with the audio.
I will try to find a link for you on this. My older ARRL Handbook has it but I guess they dropped it since mine was published.
Neil
I haven't found a good tutorial on this yet but here's an app note on amplifier testing. http://www.cleanrf.com/images/CleanRF_Manual.pdf
Though this is for a system different than what we are doing the pictures on pages 10, 11, and 12 illustrate what to look for. The "overmodulation" picture on page 13 is not what we would see for >100% since this shows a different type of distortion.
If I find something better I will post.
Neil
oh wow, thanks for all the info. now i can see that i am getting really nasty distortion. i did the transistor substitutions from that teslapress website. do you think that could have something to do with the distortion? my audio input is 1v Pk-Pk.
Probably not. It would help if you connect your scope to U1 pin 1 and then to Q9 emitter and check if you are getting a sine wave with a sine wave in at these points. If not, check the resistors around U1 for the correct values. I had a distortion problem when one of these was the wrong value.
Neil