No, you're looking at the DC component, not the 1 Khz one. The 1 kHz is clearly at -112 dB, so the -18.3 dB ia correct.The chosen presentation of the 19+20k is unfortunate - we can assume that also a difference tone is created, i.e. 1k, but the scale used totally obfuscates that range of the output - as the last x-axis scale says 5k and what the leftmost part of the scart really represent isn't readable, it's really "unlucky".
I see perhaps -105 dB 1k there to the absolute left and that aint 18.3bits - more like 17,5. Not that it matters much but in "science" we need to be correct.
//
