(When) should Crossref metadata take precedence over already existing metadata?

[cbroschinski]

Stumbled upon an interesting case today with the latest contribution from TUHH (#93). As it turned out, the file was already enriched with all relevant metadata (in the old schema, though)
Here are the first 3 entries from the original file:

institution	period	euro	doi	is_hybrid	publisher	journal_full_title	issn	issn_print	issn_electronic	indexed_in_CrossRef	pmid	pmcid	DOAJ	license_ref
Hamburg TUHH	2015	694,76	10.1155/2015/898651	FALSE	Hindawi Publishing Corporation	Advances in Fuzzy Systems	1687-711X	1687-7101	1687-711X	TRUE	NA	NA	TRUE	http://creativecommons.org/licenses/by/3.0/
Hamburg TUHH	2015	1248,46	10.3390/ma8010285	FALSE	MPID	Materials	1996-1944	NA	1996-1944	TRUE	NA	NA	TRUE	http://creativecommons.org/licenses/by/4.0/
Hamburg TUHH	2015	1874,25	10.1186/s13568-015-0122-7	FALSE	Springer Open	AMB Express	2191-0855	NA	2191-0855	TRUE	26054736	PMC4460186	TRUE	http://creativecommons.org/licenses/by/4.0/

And here is the result from the automated metadata enrichment. As you can see, there are several differences in the publisher and issn columns (journal_title also often differs, although not in these first 3). On the other hand, the original file contributes much more information in the license column, where crossref will often report "NA" values.

"institution"	"period"	"euro"	"doi"	"is_hybrid"	"publisher"	"journal_full_title"	"issn"	"issn_print"	"issn_electronic"	"license_ref"	"indexed_in_crossref"	"pmid"	"pmcid"	"ut"	"url"	"doaj"
"Hamburg TUHH"	2015	694.76	"10.1155/2015/898651"	FALSE	"Hindawi Publishing Corporation"	"Advances in Fuzzy Systems"	"1687-7101"	"1687-7101"	"1687-711X"	"http://creativecommons.org/licenses/by/3.0/"	TRUE	NA	NA	NA	NA	TRUE
"Hamburg TUHH"	2015	1248.46	"10.3390/ma8010285"	FALSE	"MDPI AG"	"Materials"	"1996-1944"	NA	"1996-1944"	NA	TRUE	NA	NA	NA	NA	TRUE
"Hamburg TUHH"	2015	1874.25	"10.1186/s13568-015-0122-7 "	FALSE	"Springer Science + Business Media"	"AMB Express"	"2191-0855"	NA	"2191-0855"	NA	TRUE	"26054736"	"PMC4460186"	NA	NA	TRUE

Question is: How should we proceed in such cases? Where should crossref (or Pubmed or DOAJ) imports overwrite existing data?
As for the current task, I decided to replace the journal_title, publisher and ISSNs with crossref imports but keep the license column.