Bostrom's reasoning about the second one is faulty, because simulations cost computational power. In a competitive environment (particularly an economic one), it is more profitable to devote computational resources to predicting the
future as opposed to peering into the past (at least to any level of detail that would give rise to consciousness). So by Bostrom's (specious) argument that there are more simulated worlds than real ones, then it's more likely that we'd be living in a simulation of the
future than the past. But since in our known past we've never had the computational resources to simulate a world to the level of detail we perceive, then it's a contradiction.
There are other, better arguments. Like, it appears that simulating our universe is computationally intractable because of the complexity of quantum mechanics. Instead it would make sense to simulate simpler worlds that don't have quantum mechanics, which of course would look completely different.
Another argument from quantum mechanics is that the outcome of most particle reactions is entirely probabilistic, i.e. relies on tons of randomness. It would be dumb to design a simulation based so thoroughly on randomness, because you cannot make conclusions based on the outcome of a single run. Instead, you'd do Monte Carlo simulations, i.e. billions up billions of runs, building a probability distribution, which would make it even more computationally intractable. That'd be completely stupid because you can't even trust your inferences without covering a sufficient amount of the probability space, and that's a problem entirely of your creation.
I am extremely skeptical of people who seriously believe that this insane reality is a simulation. They usually don't know much physics, nor computer science. A simulated world would look completely different just from the basic engineering perspective.