Joachim, It's for the reason I said.. you have to be ethically correct.. and not hide the work of others. There are people that cite their own review papers to avoid citation of others and there are people that look in one paper, find a part that is useful and cite only the citation used by those authors without knowing what is all aboutz instead of citing both. If you just use the number, cite the database. If you use the data actively (eg structure or cell parameters) then find also the original reference, check that manuscript and cite that as well so the work of everybody who substantially contributed to your achievement is recognised.
Such databases are very valuable. But they should be used as a tool to retrieve the original references (see also comments above). Note that the databases are full of errors, especially frequently if "non-standard" crystallography is involved.