Rich, there's a little confusion between distance and angle here.
Do you mean you want a 10 micron tolerance at the end of the new tool's leg, measured from a straightedge set at an absolute 45 degree angle to a baseline?
If so, and you give the length of the leg, the angular tolerance you want can be calculated.
You will obviously need some standard (no matter what method constructed) that is true to that angular tolerance.
Scribing can be extremely accurate if it is checked afterwards by reversal as mentioned earlier. If it isn't true after checking, whatever distance makes up the error can be halved and re-scribed, and then checked again. You can keep doing this until you are at the tolerance of a scribed line's thickness -- which at whatever distance out the leg extends is a tight angular tolerance indeed.
To get the last bit of tolerance possible, placing a straightedge on the scribed line, checking the workpiece against horizontal baseline straightedge. Layi another straightedge against the first angled straightedge and removie the first. This eliminates the visual aspect, and the thickness of the scribed line, and substitutes contact (which is the means of all scraping tests). Then the piece is placed against the second straightedge and checked for contact with the vertical 90 degree block edge.
When they both match side to side, by contact, it is as close as contact can measure. Which is as close as can be scraped by conventional methods.