Hello dmutz,
I can confirm your observation that there are no characterization factors present for the elementary exchanges '1,4-butanediol' and 'Cadmium, ion' for the USEtox w/o LT impact assessment method in ecoinvent v2.2. However these CF exist in the current ecoinvent v3.1 database.
In contrast, 'Mercury', does have a characterization factor in the USEtox w/o LT method, both in the old ecoinvent v2.2 database (2010) as well as in the current ecoinvent v3.1 database.
When I see such a phenomenon, I typically check the LCIA factors in my local Umberto NXT LCA installation first (I'm lucky to have both version still installed). After verifying it in Umberto, I then check whether the values I have locally are identical to the ones provided directly by ecoinvent. I typically look directly in the online database at ecoinvent.org. ecoinvent v2.2 data is still available.
Now, for the issue you addressed: Looking at USEtox w/o LT the elementary exchanges '1,4-butanediol' and 'Cadmium, ion' do not exist, while 'Mercury' exists. See attached screen grab (I have blacked out the values just in case this would be a license infringement ... although I think these CF are in the public domain).
My guess is that the USEtox w/o LT method was not correctly implemented by ecoinvent back in 2010, and that they corrected it when moving to ecoinvent v3. However, I didn't find a mention of this in the change report.
Maybe you can get a confirmation on this directly from the colleagues at ecoinvent.org.