Originally Posted by Kontact

Probably 40 from 43, but a 2mm difference is going to be difficult. The bottom of the chainring is a point pretty close to a right angle from the dropouts, but it isn't perfect, so that's a small error. (Something like 2mm for every 5° off of 90°.) And then you have the measuring error problem - did you measure to dead center of the dropout, was the fork perfectly centered. And you can refine the process and calculate out the geometry errors if you have a very repeatable measuring process. This method is just the best one I have heard that doesn't fall victim to all the minor curves and incompatible shapes that make more direct measuring methods more useful.