I’m merely proposing a mechanism for how it could be possible to have salmon return to the same spot after several generations, if that actually does happen.

The idea would be that a salmon could be genetically predisposed to follow a certain path, perhaps preferring the smell of a certain combination of chemicals, thus encoding the location. It means the „memory“ would be encoded via genetics as a result of genetic combination and mutations, and the „encoding“ would essentially just be selection. It’s just speculation on how this genetic memory idea could work without actually encoding memories on genes.

> if that actually does happen.

There is no evidence that it does.

There are cases where such mechanisms have been proposed, but postulating such a mechanism in this case when there isn't evidence of the behavior, is uninformed speculation.