App note is inaccurate about light load operation of UCC3818 PFC?

It comes down to what's "good enough". You want 10mv all-corners tolerance then buy a comparator, feed it and pay the BOM cost.

But if it's a crude backstop function that doesn't merit precision then maybe cheap is what the customer who provoked the app note (section) asked about.

Thanks, as you well know, the hfe of Q1 is massively variable, and so sizing R1 and R2 such that Q1 turns off when VAOUT falls below 1V will be a near impossible task. I know what you mean, but the cct is so poor its not even worth entertaining. I am surprised Texas even put it forward.
The circuit performs a failrly important function though, and the engineer has copied it onto the pcb we are working on......this means that we will have zero crossing noise/distortion on our pfc when the pfc operates at full power....since we cannot fit Q1 due to the above explanation.

Anyway, the circuit as per the App note is pretty useless.....so we will have to not_fit the lot of it.............because we dont want zero cross distortion when at high power........so we will have to put up with noisy operation in light load.

The UCC3818 datasheet doesnt even tell you what current can be sourced by the VAOUT pin, so you have no way of sizing the R1 and R2 resistors...would you agree?

UCC3818

I don't think this is a beta-variation problem if you put the threshold current near beta-peak Ic. It looks to be a ratio*Vbe deal and Vbe flatness is what you want (but can't have). There could be, for a particular BJT, a "magic point" (or most-so) in base network and collector load values.