I ran the circular pattern wizard and set the depth to 0.5" and 6 holes. I ran this test twice and all 6 holes both times were only 0.49" deep.

...that's .010"

Run your 3" square file in air and see if it ends at the right Z level.
That's a strange lead in with a vertical arc that I've never seen before .... but I haven't seen much. 

That is what is called a blended plunge  move .

(;-) TP

Do you think reducing the microstepping on the Z Axis driver so it takes less pulses to move 1" will have any impact on the results?

It should not make any difference unless you PC is just not up to the task.

The pulse width is set to 5us for step/dir on all axis. The TI driver has a minimum pulse width of 1.9us so this should be OK, right?

Try some experimentation here to get the smoothest motor motion (2uS gives me the best performance).

I was running the driver test and checking all frequencies. Even up to 100kHz it reports the system is OK. I noticed that up to 65kHz the CPU usage is less than 5% but 100kHz is ~50% but I don't see any reason to push it that high. When I look at Max Variation I see mostly repeatable spikes and values of 15us or 23us. This doesn't seem to change when I change the frequency but occasionally I see nice smooth line aith maybe variations of 5-8us. Is ~20us variation something to worry about? Should I break out the scope and see what the step pulses actually look like?

Unless you definitely need a higher kernel speed (to achieve the desired rapids) then use 25kHz.

Tweakie.

For completeness, the PC has Intel Core2 Duo, 1.66GHz, 2GB RAM, And running Win7 Ultimate 32-bit.

I reduced the kernel to 25kHz and the Z-Axis speed to 30in/min, Acceleration = 10 In/sec^2

You're right, my bad in saying the delta depth wasn't 0.01", it is.

I'm still not measuring a depth of 0.500". Using my Mitutoyo Calipers to measure the depth leaves a little margin of error if you don't get the plunger and caliper 90 degrees to the wood. But even allowing for a little error in the measurement technique I still don't get the results I think I should be getting with this machine.

I still need to play with the pulse width and just for completeness, change the micro stepping to reduce the number of steps. I really would have thought that the lead screw setup on the Z would have given better results than the belt and pulley setup for the X/Y but maybe not? I believe it has an Acme screw on the shaft (need to look up what I order a long time ago) so backlash should be very minimal.

The saga continues...

Well I experimented today by changing kernel speed, pulse widths, and micro stepping but I never found the magic solution. In fact now I'm consistently getting holes deeper rather than shallower like before. I'm consistently reading between 0.513-0.520" for the depths instead of 0.500". I guess I'll have to sleep on it and see if anything comes to mind.

The Z Axis in this video, https://www.youtube.com/watch?v=AfA-xQVCElg looks t be identical to what I have except I'm using a Bosch Colt instead of a Dewault.