As you can see, this does not work:
That’s the test I wanted to see! Thanks for doing it.
I have no idea whats wrong on your side but i never got such strange lightmaps, same scene like yours… No artifacts, clean shadows on my mesh.
Upload your mesh
By the way, you have 3 UV channels there, make sure that it’s not generating lightmap UV’s on import and that it has the correct UV channel selected for the lightmap coordinate. If you have a second UV channel with the lightmap UV then by default it will generate automatic UV’s to a third channel and then set that third channel as the lightmap coordnate index.
I have no idea how but i added a 3 UV channel in Max, after deleting the 3 channel… Here is the new test… by the way the standrad wraped cylinder looks excatly the same.
It’s very curious to me why your mesh looks that strange, My setup is, UV’s like you see “lightmaps”, 128 x 128 Pixels resolution, 2 Smoothing groups and all hard edges are split, lighting quality is set to production.
Ouh and the first test above with the 3 channels is not valid, becasue of the 3 channels and the pipes were set to movable.
Lightmap 128 x 128 pixels
Here is the uasset file, test it cause i’m really curious why you are getting this strange result. Warning it’s a free upload service, so there is spam.
http://www.megafileupload.com/687s/PP.uasset
After 20 to 30 different builds with different positions of the pipes , i’m getting some kind of issues but nothing strange like you have. Lightmaps are set to 128, Build quality is production.
So it works good for me, not always perfect but good.
It depends on how the mesh triangulates for how the severity is and how many segments are used for curving the mesh.
There is definitly an issue, no question about that but it seems that it really depends on the shadows or what excatly is casted on the mesh. So far i had absolut no problems but it is definitely good to know that in some cases this methode is not that optimal liki thought.
When you get into some more complex meshes it becomes too much work to try and get the small problems around the edges fixed, the issues with lightmapping are something that we’ll have to live with until we get a fully dynamic solution.
So is there any disadvantage of detaching all the elements separately and unwrapping them individually as opposed to unwrapping as one whole poly?
I could import the mesh in separate elements which would give each element a higher LM res??..
If you detach them to make separate static meshes then you would end up with more draw calls which lowers performance. In some cases you might need to split something up because it has a large surface area and you can’t get enough lightmap detail even at a high resolution.
I’m much more skilled at detaching faces than unwrapping carefully 
un-opimization FTW!