When looking at the pt, phi, eta resolutions in the embedding output processed by Rosi's (Leszek's, Anthony's...) code, I see the following:

The splitting in the phi resolution plot corresponds to electrons/positrons. Looks super weird.
This is weird and very likely wrong because in the same plots in the minimc outputs of the embedding, I see this:

(note the different y-axis ranges and different statistics).
The resolution is kind of the same for pt but the eta and the phi look very different and there is no splitting in the phi.

I need to investigate whether this is just a superficial issue (artifact of something? filling the upsTree elePhi, elePhiRc etc with weird values?) or whether this is actually serious.
That being said, I don't expect this to have a big effect on the the total reconstruction efficiency since for eta it's only a worse resolution and not a bias. And phi shouldn't really play a role.