Check every publisher, is each of them have the same fields? Maybe some of them are missing? Maybe some of them are not needed?
Some fields' data structure doesn't make sense in the parser: for example, doi (which is the only one) is added to the array. Would be enough just to take a string and join it with the "related articles" array (in generic parsing )
publication_info - Hindawi new workflows don't have the publication_info.page_start, but we have it in heprawl (however, not in other publishers in the workflow). Cannot find the exact place in the code, look like it is taken from <article-id pub-id-type="publisher-id"> and where is a pubnote
<article-id pub-id-type="publisher-id">and where is a pubnote