I checked this link. I read the article and looked at the code and test methodology. It draws conclusions about which elements are indexed and which are not, but uses the linking word "and" by itself in several areas. Search engines don't index linking words and designers never use them by themselves. So how can the experiment prove anything about whether those elements are indexed or not? Am I missing something?
The whole reason something was used there was to see whether it mattered that the space between the <a> and </a> was filled. You've brought up an excellent point though, one we didn't consider when filling that space.
Perhaps these particular instances warrant a follow-up study with an indexable word