Okay, you've confused me, I'm not even sure which of my reasoning you think is flawed, let alone why. I'll try again from the start.
You said this:
And on the other hand, as soon as a civilization is living in a simulation, it cannot create a simulation of equal complexity anymore, so as soon as it has happened, the complexity of the possible simulations drop (for simplicity, assume below 0.5 of the surrounding simulation), so withing a small number of steps (simulation-in-simulation-in-simulation...), we have that any simulated world will have a simulation complexity very close to zero.
Which seems reasonable enough, but then you said this:
As our world clearly has a complexity significantly above zero,
which isn't actually clear at all, but even if we assume that it was:
the probability of us living in a simulation is essentially zero.
...this doesn't follow. If you picked a simulation at random, the probability of it having humans in it would indeed be close to zero. However, the number of simulations that are complex enough to support humans will still vastly outnumber the number of real universes, so we're still more likely to be in a simulation than not.