mmm....
I think the equation in the datasheet "(TI) log112" could possibly be wrong, there is NOT "negative" relation between Vlogout & Vo3.
But the simulation shows they are "negative", possibly in the "k" factor.
"VLOGOUT = (0.5V)LOG (I1/I2)"
"VO3 = K (0.5V)LOG (I1/I2), K = 1 + R2/R1"
Maybe someone should inform (TI) to fix this. If it's really wrong or there is a better explanation.
Best regards.