So I am slowly going nuts. Here's the story, a continuation of an earlier saga:

I ordered some Ambrosio Nemesis rims. Of course I needed spokes for them, so I looked up the ERD. It is commonly thought to be 617 according to various internet sources (including spocalc). I also asked the shop where I bought the spokes--they told me the ERD is 614. I received the rims a few days ago and went to my LBS with them. They measured using some rim rods and told me that the ERD is 614.

But wait, there's more! I just sat down and measured them myself. Using the method where you thread nipples onto two spokes, pull them taut and measure the distance between the nipple ends then add twice the nipple length, I got an ERD of 617.

Then I learned about the way Jobst Brandt apparently measures rims. He sticks a spoke into a rim hole that already has a nipple seated in it and measures the distance between the spoke head and the outside of the rim, then measures the diameter of the rim--outside to outside-- and subtracts twice the first distance. Guess what: the ERD using this measuring scheme was 614.

That is, until I learned that ERD is measured to the bottom of the screwdriver groove on a nipple, meaning that using Jobst Brandt's method I should actually add 1.5 mm per nipple back (this is the measured depth of the groove on the nipples I am using), but using the first method it seems that I should be adding only 10.5 mm rather than 12 (though an illustration of this method shows that you should add the whole nipple). This simply reverses the measurements--Brandt's gives me 617 and the first method gives me 614.* edit: this paragraph is actually wrong. See end of post.

For a certain ERD length, is spoke length calculated as to be flush with this groove taking into account that people measure ERD encompassing the whole nipple? A cursory glance at the spocalc formula leads me to believe that there is no such correction. Based on these exercises it seems to me that I am in the ballpark and it doesn't really matter if I ordered spokes based on an ERD of 614 or 617. The LBS told me not to sweat it, and upon emailing the show I bought the spokes from they said the same thing. On the other hand, people on the internet seem to think--contrary to these shops and Sheldon Brown--that if spoke length is off by more than 1 mm from the calculated measurements, you are screwed.

Another idea is that I should measure the ERD according to the instructions of the spoke length calculator I'd be using. In this case, the instructions on Damon Rinard's spocalc page (the two spokes pulled taut + length of whole nipple) are right on the money. If anyone actually reads all this, please share your thoughts!

edit: actually, in my caffeinated and clear thinking state, I now realize that I am wrong about "correcting" Brandt's method. I would have to subtract 3 more mm if I had measured into the groove of the nipple.