To help to demonstrate this point, I just carried out a little experiment. I took a wheel that nominally accepts 296mm spokes. I fitted three spokes in turn to the same (outside) position in the wheel, each using a slightly different approach, and tensioned them. The crossings in this wheel were not braced, so that was one less variable to worry about.
A new DT PG spoke, fitted in the usual way, and then enthusiastically stress-relieved, came flush with the top of the nipple
A nice new DT PG spoke, when tensioned but not stress relieved, came flush with the bottom of the nipple slot..
A third new spoke was spoke washered so that so that the J bend was a snug fit and needed to be cold-set by a few degrees (which BTW constitutes the best fit, IMHO). When tensioned, this spoke came short of the bottom of the nipple slot, by about two turns.
The photograph below shows the three J-bends in order, from top right to bottom left.
- elbows out...
The middle spoke is still in the as-received shape, and represents what happens to many spokes that are fitted into hubs with a loose-ish fit, in wheels that are not stress-relieved; they come out looking the same as when they went in.
[ IME this is unsatisfactory, in that the spokes will see considerably greater fatigue loading in service; either of the other two conditions is an improvement over this.] The bends in the other two spokes are of slightly different shape only, but this was enough to make them about four turns (~2mm) different in length in the nipple.
Note that this experiment was carried out using a steel hub, which (through use) had a slight radius where the inside of the J-bend bears against the edge of the hole. This was unchanged after the experiment, as judged by refitting the second spoke. Note also that most decent wheels are built onto hubs with aluminium flanges, which have considerable potential for deformation (in service and/or during stress relief), and this creates yet further scope for altering the apparent spoke length. Had the experiment been repeated with 15g or 13g spokes I daresay there would have been a different result.
Thus the idea that you can just ignore the quality of fit of spokes into flanges is clearly not one that holds up under scrutiny. Whilst it is true that if you always use the same types of spokes in the same type of hubs for all your builds, you may not see much variation, few wheel builders will always be able to do this, and some attention needs to be paid to this aspect of the wheel build.
In particular if you use spokes with different J-bend lengths, and/or improve the fit of spokes by using spoke washers where necessary, spokes that are (without washers) 'just right' for length (but not for fit into the hub) are liable to come up a bit short in the nipple once washers are fitted.
After inaccuracies in ERD measurement, IMHO this is often the largest source of error when calculating spoke lengths; even without differences in the way the hub settles, it can produce variations in apparent spoke length of ~2mm.
cheers
