Backlash does not repeat - by that I mean that doing it 10 times or 100 times makes no difference, the backlash is always the same.
It is the amount the axis "dwells" at either end of a move, as it changes direction. If you set off from zero, and move 1 inch, that will be accurate (if you moved to zero in the same direction. If you now move back to zero, the axis will now come up short, becasue of the backlash the axis had, whilst reversing.
However, if you now move again to the 1 position, this will be accurate again (because the axis did not start from zero, and backlash could be taken up before the axis moved) - so backlash only counts once, at either end of the cut.
You cannot be loosing steps, becasue you say that one of your axis moves too far. This must then be steps per unit.
You must CALCULATE the number of steps per unit - not measure.
If your motors are steppers, they need 200 pulses per rev.If your drivers have microsteps, then your must multiply by that - e.g. Gecko is 10 - i,e, 2000 steps. Your gearing, through the timing pulleys, must be taken into account - I have 3 to 1 on my lathe - i.e. 6000, so you come to the final drive.
I do not understand what your final drive is, on an 8' x 4' gantry, but this is simple enough - how far does it move for one turn of the drive spindle from your last timing pulley. From that you must work out the number of pulses per inch - or per mm.
Be careful with your final drive, is it imperial or metric - a lot of stuff these days comes in either, and is wrongly marketed as being imperial when it is metric and vica versa.The conversion is 25.4mm to one inch - and the .4 makes all the difference
You say your carriage moves exactly one inch, when you move it one inch. Make sure you are taking any backlash out by - moving right - stop - set up measuring device - move right again - stop - measure. If this is exact, then move back one step and see how much short this is on the return. That is your backlash.
You can enter this in the backlash table under Config and turn on the backlash compensation.
The only other thing we have had on these pages is that some of the electronics (breakout boards particularly) which inverted signals, were sometimes prone to miss the first pulses, which means that things came up short, bu I wouldn't bother with that till you have checked everyting else.
Sorry I have gone on a bit, you probably know most of this anyway.