There are two ways ERD goes wrong.
First there are variances in rim manufacturing. Some manufacturers are better than others and to some extent you get what you pay for. But I always measure and frequently find rims out of spec or find differences between pairs of identical rims. If you know what you’re working with, variances are no big deal. If you input correct dimensions, you’ll get correct spoke lengths.
The second issue is differing definitions of ERD. Typical definitions include penetration of the spoke into the nipple head, which is important for maximum strength. But how far the spoke extends beyond the nipple seat (into the rim) is unclear. It depends on nipple geometry and encodes some preference for how much penetration is ideal. In other words ERD is relative to an undisclosed nipple and opinions regarding its use.
Typically the magnitude of these errors is pretty small, maybe 0.5mm each. Additional error is introduced when you round spoke lengths to match availability (why buying spokes in 2mm increments is rough). Maybe these errors add up to a problem and maybe they cancel each other out.
This sounds like a pain but it’s not at all. You simply make your own tools and take your own measurements. The parts cost next to nothing, assembly takes 10 minutes and you’re left with tools you can use forever. Maybe it seems like a pain because you want your spokes and rims to arrive at the same time. That’s fair but getting it right the first time is always fastest.