[PHI 3000] Is auto-epistemic reasoning reliable?

Auto-epistemic (self-knowledge) reasoning is inference from premises about what we actually know to a conclusion about what we would know if something were the case. For example:

1. If Pegasus were real, then we would have ample evidence for its existence by now.
2. We don't have ample evidence for the existence of Pegasus.
3. Therefore, Pegasus does not exist.

Is this kind of reasoning reliable (i.e., does it lead to true conclusions more often than not)? It seems that we can think of instances in which this sort of reasoning would fail us. For instance, perhaps the evidence is very difficult to come by, as in the following example:

1. If there were extraterrestrial life, then we would have ample evidence for its existence by now.
2. We don't have ample evidence for the existence of extraterrestrial life.
3. Therefore, there is no extraterrestrial life.

Or perhaps it is not clear how long we should wait until we can say with sufficient confidence that, if no evidence has come forth until now, it is unlikely that such evidence will come to light in the future, as in the following example:

1. If there were atoms, then we would have ample evidence for their existence by now. [Said before Einstein's work on Brownian motion.]
2. We don't have ample evidence for the existence of atoms. [Said before Einstein's work on Brownian motion.]
3. Therefore, atoms do not exist. 

When is it reasonable to rely on auto-epistemic reasoning of this sort?