Surface / grid probing with the Centroid Acorn

Flatness problem:
I've got a moulded plastic base filled with potting compound that forms the bottom of a product we are developing at work. The product (an SMPS) relies to some extent (in some applications) on conduction cooling through the base into the chassis or heatsink it is mounted on, so a degree of flatness is helpful unless you want to throw loads of money and thermal grease at it. The moulded base suffers some distortion during the cooling / solidifying stage and the potting compound also undergoes shrinkage as it cures, so there are several effects working against us.

We did a scan on a CMM at work but ran out of time for the moment. What better opportunity to check out the digitising function on the Acorn?

Centroid Digitise function:
Setup is quite easy and almost intuitive:

The "patch" length and width are pretty obvious, as are the stepovers. The "Z step up" is how much the tool lifts after touching and the "Z maximum depth" is how far below the start height the probe is allowed to seek downwards when looking for the surface.

With the values above, I should get a 127 x 77 matrix of measurements ie 9779 data points on the face of it. As it takes about 1s per position, that's about 160 mins ie 2 hours and 40 mins. Sure enough it took over 2 hours, while we were out shopping etc.

Control surface:
The probe may be giving me bollocks (inconsistent) readings, so I also need to measure a control surface, such as a precision ground parallel. If that genuinely flat surface gives massively noisy results, I can take the actual measurements with an appropriate pinch of salt. From what I recall, my backlash is around 10-20um. However, as these measurements are all taken in the same direction they should be largely free of the effects of backlash, so I'm vaguely hopeful the measurement "noise" may be better than that. Time will tell, all 160 minutes of this - plus another hour or so of probing the control surface. There's really no need to use the same tiny stepovers for the control. Or if you like, there's little point scanning the whole surface. I could repeatedly probe the same X/Y position to the same effect.

Now to analyse the data. Or more to the point, HOW to analyse the data?