This question somewhat illustrates how using tools like the DT Spoke Calculator and similar can make things more complicated and ambiguous than they have to be, owing to their attempts at hiding some of the technicalities to appear more user-friendly.
The answer is that it depends on the threaded length of the 16mm nipples in question. I believe most 16mm nipples have around 1mm longer threaded section than a 12mm, but I don't have a DT 16mm in front of me to check. I also believe not all 16mm nipples are the same in this regard. And it also depends on whether and how the DT calculator is pre-compensating for spoke stretch and rim compression. (Namely whether that "accurate" number is already attempting to reflect it.)
Here's how I would approach this. Use the free spreadsheet Spocalc or other software where you can see for certain that the formula being used to spit out the numbers is the classic spoke length calculation formula in its simplest form, unadulterated by anyone's built-in fudge factors for nipple type, spoke stretch, and rim compression. Input an effective rim diameter (ERD) that you measure yourself with the nipples in question (you can do this easily with 2 spokes in the ballpark length, 2 nipples, and a vernier caliper).
If it's unclear, here's why you would do that. ERD is typically defined as the circle formed by the ends of the spokes at their optimal thread engagement, which usually is itself defined as the spoke tips being flush with the top of the nipple. Some combinations of 16mm nipples and spokes (which also vary in their threaded length) won't let you do that; you run out of thread with something like 1mm left to go before the spoke tip gets flush with the top of the nipple. In other words it's topping out around the base of the screwdriver slot, not the top of the nipple. If you used that combination of spoke and nipple while still using lengths generated with an ERD measurement that had the spoke flush with the top of the nipple, you could easily run out of thread and not be able to properly tension the wheel, which is the situation DT is trying to guide people to avoiding by offering the ability to switch nipple types in their calculator.
The length numbers that Spocalc or similar gives you are what I call the "pure geometry" number. From there you need to compensate for spoke elongation and rim compression. I've learned my take on the numbers for this from the shareware software Machinehead Wheelcalc, which you can download and check out, but for practical purposes I use around .6mm for a 2.0 and .8 for a 1.8 if I expect the tension to be in the 100-120kgf range (dishless wheels and drive/disc sides of dished wheels), or about .2mm less for both cases if I expect it to be in the 60-70kgf range (the slack side of most dished wheels basically). For 1.5mm spokes at high tensions (front wheels basically because I don't use them on drive side rears) the number gets all the way up to 1mm, or .5mm for left side rear.
From there you round down to the next available spoke length. The most I go less than that number is 2mm, and if I'm using aluminum nipples I'd be looking for ways to keep it to 1mm or less. Brass nipples let you cheat here a little, maybe by another 1mm, if needed or if something goes wrong, but al nipples really don't.
