Do you have images of the latter two results? Them all behaving so similarly suggests that you're not exactly getting the amplifier itself. Do you have almost any opamp around that you could wire up as a unity gain buffer to "go"/"no go" as far as measurements on that?
Yeah, that's what I wonder as well. How do the amps behave at roughly 1 watt into 4 ohms or less? That will show fairly clear glitches from under bias.
