It was auto-correct, actually.
Also the rest of what you're saying is incorrect. With buffer, this graph shows distortion dropping below -100dB only at 500mw. Without buffer, the level is much lower at 20mw.
However this is at a single frequency (probably 1khz). If you take a peak at these two graphs, you get a much different story.
If you review these, we see a much different story, to the extent that it almost proves the first graph entirely wrong, given that this graph shows no tested tone dropping below -100dB SINAD until 3w w/buffer, 500mw w/o buffer. We also see that certain frequencies /never/ drop below ~-97dB SINAD. To make matters worse, these impact primarily higher frequencies. The human "ear" is more sensitive to distortion at higher frequencies than lower ones, and the two primary distortion culprits here (10khz and 15khz) directly impact air/stage. Distortion at these frequencies tend to lead to poor imaging, stage, and a sense of clutter.
Given this distortion, I have grave concerns about the linearity of the amplifier under load real-world. At current, Amir doesn't run any linearity tests at real-world power levels (5w is abysmal for speakers) or against loads that mimic real-world load impedances.