Zero has found the document and has posted a very
helpful analysis at
http://en.wikipedia.org/wiki/Talk:Estimates_of_the_Palestinian_Refugee_flig…
In summary, the 472,000 figure was indeed in the document, but Zero
feels Bard may not have interpreted it accurately.
I agree that citing the original document is better than citing a
secondary source, though others might argue that makes it hard for the
reader and other editors to check that Wikipedia is quoting
accurately.
I do hope that nobody would make such an argument. We should always cite
primary sources where at all possible, and this instance shows the
importance of correctly handling secondary sources. It isn't so much a
matter of a secondary source being intrinsically unreliable, it's the
danger that we may replicate possible errors in secondary sources by
citing an attribution to primary source as factual when, not having
checked it, we do not know it to be factual.