I struggled with ball size for a while! The original probe was 3mm, made in one piece with the shank.So I tried to replicate it by fixing a 3mm ball to the end of the probe. Epoxy and LocTite 603 "retaining compound" failed to reliably retain it (603 used because I happened to have some). Then I realised that 3mm was a bit arbitrary so I turned a ball on the end of a bit of mild steel (knew I would find a use for my ball-turning tool eventually). Ended up about 4.78mm or something - but working to an exact size doesn't really matter as it is just a parameter set in the probing routines. It is also very difficult to adjust the thing as almost the smallest perceptible tweak of the adjusting screws shifts the probe quite a lot. However, while repeatability on a single axis is fairly good, if I rotate the thing I get different results. I suspect that it's an issue with spring strength as well as the issue that the force needed to break the contact depends on the geometry of the pins. I'll have one more go at it and then probably accept the fact that it is cheap and cheerful and for most of my purposes, adequate.

I did do a search for this info and found a long thread on CNCzone started by someone who was building a probe to measure to better than 0.02mm. He was concerned about things like thermal expansion of the materials involved and did a lot of simulation as his workshop temperature could vary a fair bit. Then he said that, for example, he needed to probe a car dashboard which might take a day or two (hence concern about temperature variation) but there was no mention of compensating for expansion of the thing being measured with respect to the rest of the kit. When you are searching for real precision, you really have to look at every detail which might be why a Renishaw probe could cost thousands.

On the subject of wigglers - I still use one on my manual mill. And very good they are too - although thinking about it, about as repeatable as my touch probe! Mind you, I am on my second wiggler after starting the mill to touch off a piece of work and not realising that it was set to 2K RPM. Probably did about 3 revs before the probe swung out, hit the work, and the probe snapped where it met the pivot ball. Not sure the probe didn't ping off the other side of the workshop... I now have a sliding-disc type rather than the ball-on-stick type.