spans indicating <b>, <i> and their friends
At least one known target (Editoria) is not ready to deal with arbitrary spans with implicit or unknown semantics.
Due to the way data extraction is happening, we should never see <span style="font-weight:bold">
only <b>
- so a span of that sort is not likely to be lost. However, it is marginally possible that further down a pipeline, such elements could be created, for example by some sort of ad-hoc mapping or cleanup -- and those will not make it through a final conversion into the stricter tag set.
We might consider converting clear cases to inline tagging <b>
<i>
etc. as a preventive measure against info loss in such cases. (Or simply be vigilant.)