I think I know how to fix the over extrusion issue. So I go to C:\Program Files (x86)\MakerBot\MakerWare\default_configs and find replicatordual.json, you may need to open replicator2x.json for your Makerbot Replicator 2X. And I open the file with notepad++, you can use any text editor to open it. and I changed steps_per_mm for A and B from -96.275 to -88, which greatly reduced the over extrusion problem. I can then fine tune the extrusion by adjust the filament diameter.

I try to search where I can adjust step motor driver current, with no success with any of the keyword I tried, so I went through all code, and find "G130 X20 Y20 A20 B20 (Lower stepper Vrefs while heating)" interesting. It seems like makerbot use G130 to control step motor driver current. So I search all G130 in the document, and found "G130 X127 Y127 A127 B127 (Set Stepper motor Vref to defaults)" under end_start_sequence, so I change it to "G130 X100 Y100 A100 B100 (Set Stepper motor Vref to defaults)", OMG, the noise is so much better, I probably can lower the number further but this already make huge difference.

Keep in mind after you edit this file, you need to restart makerware and redo the slicing, so use a simple model to do the test, otherwise it is going to take long time. I'm not sure 100 is the best value for motor Vref, I need to run more test, but I imagine the best value will be machine specific, so it is best you do your own test.

You can also adjust acceleration (max_speed_change_mm_per_s) and jerk (rate_mm_per_s_sq). I also changed        "max_inner_shell_speed": 90, to 40 so that it is same speed as "max_outer_shell_speed". There are also many other setting can be changed, and I'm quite happy with the printers performance after preliminary adjustment.

In my experience, such small difference can be fine tuned by adjust your flow rate, by reduce extrusion slightly, your part will become slightly smaller (thinner) after cool down,

printed part to printed part should not be an issue - printed part to machined part, yeah, it can be problematic.

The best way I can suggest to cope with the minor issue is to compensate in the modeling stage.. If you know it is going to be K.K mms off in
B direction, then you adjust your model that amount in the correct direction to offset the tolerance difference.

If you really want to tune it do a series of circle and determine if the error grows . If the error does not grow then you have a backlash issue.  if the error is doubled on a double size circle then you can tune your steps. also it is easier if the error is aligned with one axis . if the error is at an angle you may  to play with some trig.

Makerbots do not run firmware like Marlin or Repetier. There is no open source to the firmware you can edit and upload. The closest thing is Sailfish and even it has limited editing. My CTC is a clone of the Makerbot and their firmware limitations was one of the reasons I swapped out my logic for the Azteeg X3 I use now. In fact I still have the original board and display.

Thanks
Dale

Thanks Carl
I was going to try and get it closer because parts are not fitting together.

As long as all the parts that fit are printed they should fit. The issue of tolerance comes into play when you are trying to fit printed parts to machined parts. Like in my case I am trying to fit
machined rifle barrels to printed parts and they are sometimes tricky. Don't ask......

Thanks
Dale