All Glowforges do not have exact, 0.000" correspondence from lid camera bed image to design placement to resulting cut/engrave/score on the material over 100% of the honeycomb area of the crumb tray.
There are variations experienced among many Glowforge production units, variations through time from the first PRUs almost a year ago, and variations according to use cases in individual Glowforges. @Tom_A got a unit swapped because it was found that his variations were not within acceptable (according to Glowforge) limits. Several others have too.
Some units have been improved remotely by whatever tweaking is possible from the mother ship.
The 1/4" tolerance is not a standard that can be applied equally to everyone’s experience. some experience much better position accuracy and some don’t. I am unaware of any consistent research available to indicate the different kinds of variations among different users. Once way back when we were trying to figure out whether or not our lid camera images were sharp or blurry and what to expect as a standard, many of us exchanged lid camera images to help. It was discovered that there were improvements to be made in both lid camera fixture and software issues. I do not remember if anyone had to get a new unit because the lid camera image could not be clarified remotely and was just too blurry to use.
I will post my real world example today of cutting a circular object that is 2.5" x 2.65"
The location on the acrylic, non-proofgrade Chemcast .215" thick and entering material thickness as that in the GFUI is a narrow band that is 2.55 inches wide at its narrowest but with the cutouts, I can fit this object into the spot. I expect that the Glowforge should cut this out. That means that I can count on close to kerf width accuracy in placement.
Kind of hard to see the light teal circle at this zoom level, but it’s visible.
I’m writing this as I perform this operation. Let’s see what happens.
And iPad shot after.