Hey guys,
Here's a little update on where I am at. In short, I have no clue what this mysterious difference is between the 2K18 and 2K19 models that is causing the issue. I don't need to know, though - I have been analyzing the binary (model) file and mapping out its structure so that I understand how and where the important data is stored. I think I can copy the face/head data of the converted player into the 2K19 player's file in a hex editor. This is not guess work either, but something that can be determined down to the exact byte at the beginning and end of the selection.
The goal is to have the head of the converted player and the body of a 2K19 model, so that the issues with bone assignments would not be an issue. But the nice thing is that despite some differences in geometry in other parts of the model, the heads of players have the same # of vertices and the same triangle count in 2K18 and 2K19. This means that they occupy the same amount of space in the file, so that while transferring the data we are not messing with the surrounding format in the file.
I don't want to make any promises but this has a chance to work. Regardless of how this turns out, things will be pretty concrete as to whether conversions can or can't be done this year. I can't really dig any deeper or go any further into the rabbit hole because I am examining the raw binary data in individual bytes for the model - there is nothing more to be done that I am aware of.
As you can see in the picture above, this is the converted player's head with the rest of it being the 2K19 model. This probably looks similar to some of the results you have seen when we attempt to combine the models in blender and then export them.
The difference here is that this is the result of making changes to the file in a hex editor. The face is obviously still messed up but this result is only from copying in the vertex coordinates of the mesh of the head. In this case, the position of the vertices changed but it still looks messed up because there is more data associated with the vertices that needs to be copied. Besides that, notice how from the hex editing, no other part of the model has been disturbed. The skeleton imports and the bone assignments for the vertices are correct.
Right now I can manipulate a single vertex with a specific vertex ID # and spot exactly where in the file that change takes place. I know the sections where this vertex data is stored. But the next step is to pinpoint where the other vertex data is within those sections, so that I can get the data I need while keeping the vertex ID's the same and avoid disturbing the data for the rest of the model.