There' is logic to what you profess. Neck tension is the key. Brass with uneven neck thickness would not be considered for my needs . I am looking at the collect crimp as a solution to consistent neck tension with out adding bullet run out. I use the Redding competition sizer and seater. . I found that the sizer did more to reduce bullet run out than the seater. Especially when necking 280 down to 6.5. I like to leave a short section of the neck unsized to insure bullet alignment to the bore and just barely kiss the lands. I can only do that with bushing dies. If your chamber is dead true this is I believe is the best way. The only neck size collet dies I am aware of are lee. Any body else produce them. Does t the bullet push any uneven neck thickness to the outside just like a mandrel? Are you measuring run out
The problem with using bushing dies to only size a portion of the neck is that you can run into the same issue I spoke about, as far as not enough uniform neck surface in contact with the bullet. The rule of thumb is no less than the the diameter of the bullet. As in, if you're using .308" bullets, you would want no less than .308" of the neck surface in contact with the bearing surface of the bullet. That is what is key in maintaining minimal bullet runout.
I have used high dollar sizing dies and had good luck with them. I've used the Lee collet neck sizing dies too. I've used the Redding Competition set that has a body die, a micrometer bushing neck sizing die, and a micrometer competition bullet seater. Reluctantly, my testing proved that the cheap Lee full length sizing die with the expander ball removed, in conjunction with a mandrel die, produced the absolute best and most consistent ammo with under .002" runout. My high dollar sizing dies sit and collect dust anymore.
As far as the bullet acting as a mandrel as it seats, yes, it sort of works that way, but if the necks of the cases have become work hardened, they will have spring back. The amount of spring back will vary depending on the thickness of the brass. Thicker areas will have more, and thus produce higher tension at those points on the bullet. This is why annealing and neck turning are part of my process. It undoes the work hardening and ensures uniform thickness. Using a mandrel, and then seating the bullet still seems to be more effective than a bushing die- without neck turning as well.
This has been my experience. I've discussed this with a few others and they've experienced the same things. No one wants to admit that after they spent a lot of money the results are still inferior. Sometimes guys refuse to see the truth as a result. Sometimes, results are indeed different or minimal enough that a conclusion is hard to make.