I don't have access to files now but basically, follow the directions we gave earlier.
Step 1, edit the calibration script by copying the z home line 3 times. You must do this or you'll never hit the endstop.
This records correct offsets
Step2 edit the latest and greatest cupcake profile from rep-g 26. Later versions do not have good cupcake profiles, older versions have a change in the way homing was done. Previously, homing offsets were NOT stored anywhere, the start.gcode was edited to match your hardware.
This is why you are confused as to starting at 0,0,0. Technically, the old cupcake (without endstops) also started at 0,0,0 but you did this manually every time you printed before there was homing. The offset method now is we run to the endstop, we assign the offset saved value, then them move back to 0,0,0 from the endstops and thus it's the right place (provided you calibrated right, and didn't run into the timeout homing). This is why you MUST edit both the calibration script and the start.gcode.
Also, while on the subject, have you not discovered the ABP sucks? You can probably print 10 things before the belt gets wrinkles in it, and worse, you will have problems building anything taller the the calibration cube. Take your heater plate and just make an HBP with kapton tape.
Further, the extra weight of the ABP and the weak cupcake stock motors will cause you to skip steps and ruin prints at some point. It's not worth the headache. There is no source of ABP belts, so you would have to cut your own from laser printer overhead slide sheets (don't use ones for inkjet as they will melt and gum up the rollers). Just because Makerbot got a patent for the ABP doesn't mean it's worth using.
Again, I'll take some time and type up the files, but not worth using in all honesty.