What settlers have really seen also depends on whether your prior is correct in our world of real islands (implicitly 1/4 to each possibility). But you can easily model what people would see in a world with that distribution of islands:

e.g. Let an easy to get to island have 100 visitors and a hard to get to one 10 visitors. Let visitors to disaster prone islands die before the next people arrive. Let there be ten of each type of island

Total observers:

From each island type 1: 1 arrives to an uninhabited island, 99 to an inhabited one

Island type 2: 1 arrives to an uninhabited island, 9 to an inhabited one

Island type 3: 100 arrive to uninhabited island

Island type 4: 10 arrive to an uninhabited island

Of the 1120 observers arriving at uninhabited islands (ten of each of the above islands), 1000 will be on type 3 islands, consistent with SIA. In this case the alternative anthropic principle preferred by others, the self sampling assumption (SSA), can come to the same conclusion because the possible islands you are considering are actual. If real islanders have seen something other than this, it is because there were different frequencies of islands.