Let's try that again - and see how it actually measures up

Having been away from the workshop on a family break, I've been itching to run off another example, with a few changes / enhancements made for good measure:

• Increase the threading speed. Currently 500rpm but it seemed happy when running over 1000rpm - I had no programmed dwell between the 2000rpm turning spindle speed and the lower 500rpm speed used in the subsequent threading operation, yet it seemed to cope fine during the initial phase where the spindle was still decelerating.
• Slight increase in diameter of the taper. This is a very shallow taper and although the finished dimension seems close to tolerance, there is little visible clearance between the axial faces of the taper and collet when assembled. So a more finite / visible gap would be reassuring.
• Use a VNGT (polished and honed) insert for the roughing and turning operation, to see if I can get a better surface finish. These are often described as being "aluminium cutting" inserts but some are actually specified as also being suitable for medium to light machining of steel. True, they have no coating but buildup and chip welding doesn't seem to be an issue for light cuts.

First attempt - went fine, if you don't mind a thread that looks like a serrated weapon. The threading was very noisy, so this was due to chatter. I had the tailstock supporting the workpiece, so I concluded that my surface speed was the issue. I can't sensibly try to operate at higher speed (often this is the suggestion to get past chatter), so the solution is presumably to reduce to perhaps 1000rpm.

Let's try again, ideally not stopping the recording part way through, requiring a 2 part video:

That's better. 

Can't see it easily here but it looks to me that there is a small but finite gap between the collet and the adaptor once it's tightened:

But the bottom line is - have I improved the runout at all? Well, not massively. I'm still seeing about 100um (0.1mm) total runout, which is a lot for a small cutter. 

Not massively happy with this outcome. On the upside, it's an improvement on the previous runout (250um / 0.25mm) but it's still pretty miserable.

I will need to look more closely at the assembly, to figure out where the runout is concentrated. Is it the adaptor or have I overlooked something? Could there be insufficient clearance between the collet and adaptor? Is the centre pip in the Autolock collet chuck off centre?