Wait ... what? Whoa. I wonder how many others have misunderstood this like I did. I did watch the HLN special, and even then I didn't pick up on the fact that this is supposed to be two separate people. I mean obviously most of us thought they looked nothing alike, but even in the special a relative was describing how she was able to reconcile the two as having the same features just described by different people.
The wording in these bullet points makes it sound like Sketch #1 was someone else entirely, perhaps a specific person someone had seen at the park and that maybe they even know who he is and have now ruled him out. And Sketch #2 is supposed to represent BG. Which is super odd, because it looks nothing like what we can see of the video, and the age range for the voice wouldn't match up, yet Sketch #1 does seem to represent BG. What on earth?
I have no idea why LE in this case chose to use the terminology "they are two different people." It would have been much less confusing to say, the sketches are two totally different recollections of an event (the event being, seeing someone who in retrospect was suspicious). Once you stop thinking of each sketch as a verified "person," and more "someone's memory of a fleeting moment in time," it becomes easier to understand.
Think about how composite sketches are created. We know the "young" sketch was created from the memory of a woman who "saw something she felt needed to be reported" and it was drawn just a couple of days after the murders. The trooper who drew this sketch told us his process. He sits down with a witness and asks them to look at a bunch of different chins. Then the witness selects the one that supposedly best matches the memory of what they saw. Then the witness looks at different sets of lips and chooses. Then noses. And so on. This is how a composite - a matching together of different parts - is created.
Recent research has shown that we should perhaps question if this piecemeal approach is really the way that the human brain sees and remembers faces. Human memory, especially under stress or if you didn't realize a moment was significant, can be faulty too. But no matter - a composite is only as accurate as a witness's memory of a face.
Compare that to a surveillance photograph of a POI in a hypothetical crime. Let's say, one taken from a camera at an ATM. It's not a memory. It's a physical recording of a real person's face at one moment in time, exactly the way it really looks. Would you not say that, all things being equal (face covering the same, etc) that the photograph is more accurate than any composite sketch of the same POI? Ask yourself why it's more accurate. Because nothing about a photograph is made up based on imperfect recall.
This is why Carter has said that a sketch is not a photograph. It's not a real, verified person who exists, like the surveillance image is. It's just a collection of facial features recalled by a witness, perhaps not that accurately. Carter knows that the person, when found, is unlikely to resemble either sketch to a great extent. That's why he said that the killer may look like a combination of the two sketches.
Even if the two sets of witnesses (a single woman for the young sketch, two or more witnesses for the old sketch) were describing their memories of the same person, they are remembering different aspects just based on what they noticed at the time or could recall later. This is what BP was getting at in the HLN show - witness 1 may have remembered a young person compared to her own experience/viewpoint and the group of witnesses that contributed to sketch 2 described a different memory based on their viewpoint/experience.
Clearly LE have decided based on evidence that one memory is more accurate than the other. Because the memories of two sets of witnesses created two very different looking recollections, one became secondary. Because the sketches aren't real people, this doesn't mean anyone was cleared.
So what is the point of a composite sketch if it's not a "real" person? The hope is that there may be one or two recalled features that are accurate enough where they would stand out to someone who actually knows the person well and can identify them - not just based on resemblance to the sketch though, but by the totality of their behavior. This potential witness may see the sketch and think, it doesn't look exactly like Bob, but the nose does look like his, and he was off work that day, and shaved his beard and hair the next time we saw him, and he knows those trails, and so on.
This is why LE in this case keep asking people not to do side by side comparisons. They know this person is not going to be found by strangers going through old Delphi yearbooks. Sketches work when someone who really knows the subject can put together the resemblance plus behavior.
Sorry for the lengthy explanation but I hope it helps someone understand who gets this far.