Thanks for the report @ShaunMcD !
This is my fault. We have an issue for this https://github.com/stan-dev/loo/issues/85, but I forgot to fix this after this was fixed in monitor.R. I’ll fix this tomorrow by copying the updated code from monitor.R, but it will take some time before the next CRAN release of loo package (we should fix some other open issues, too).
There’s a (longish) discussion of the differences in N_eff BDA3 vs. Stan