There are other studies like
this one from Iceland that show less than 10% did not seroconvert after natural infection and t
his one from the UK showed 8.5% nonseroconversion. But we do know that this number does not stay consistent as time progresses post-infection.
This study suggests that, "Especially in
asymptomatic cases, antibody responses disappeared after 8 weeks in 40% of individuals in that study." (emphasis mine)
So it's difficult to estimate due to the noted effect above, but we can use ~20% that were infected and didn't have detectable antibodies (that doesn't mean that their memory B-cells wouldn't produce them upon reinfection, but that's a different topic).
As for the vaccinated, a much larger number seroconvert as noted by
@The Football Freak, but that may too fade over time. However, that Rueters article said this, "The survey showed 67.6% of adults were seropositive, while more than 62% of adults were unvaccinated. As of July, just over 8% of eligible adult Indians had received two vaccine doses." We know that seroconversion happens after the first dose, so we can use 35% of the population having vaccine induced antibodies. Now there's also the crossover folks that were infected naturally AND received a vaccine dose. I'm not sure how to get at that number.
Even if 10% of that 35% didn't seroconvert, that's still 31%, so higher than the 20% as noted above. Therefore, these numbers likely don't wash each other out. But my larger point is that math of 2/3rds of 1.3 billion is way too simple, as is epidemic modeling. Both natural and vaccine induced responses (along with NPIs) contributed to the slowdown in Delta in India. It's rarely an either-or scenario.