Older forms found on LC bib records raise a number of questions. Click "Continue reading ..." below for an extended discussion which covers a number of them.
For 670s generally, you cite the source you saw. When a 670 cites "LC manual cat.", it means that someone at LC actually looked in their old card catalog. If you found information by looking in the online LC catalog (which I did for the example used here), that can be reported as follows:
670 LC database, May 2, 2006 $b (hdg.: Fisher, Warren Samuel, 1878- ; usage: W.S. Fisher)
Note that we're citing all the information we found in the file related to the heading, not just a single record. (Source for this in Cataloger's Desktop: Descriptive Cataloging Manual: Public Sections/Z1. Name and Series Authority Records/Variable Data Fields/670 Source Data Found/Special types of citations/LC database.)
The first 670 in an authority record should be for the work being cataloged. In this example, that work shows "Warren S. Fisher." That usage form combined with the additional information found in the LC database and constructed according to AACR2 gives the heading form
100 1 Fisher, Warren S. $q (Warren Samuel), $d b. 1878
The usage information in the 670 above would justify an added 400:
400 1 Fisher, W. S.$$q(Warren Samuel),$$db. 1878
LC would not make a reference for the unestablished "old catalog" heading "Fisher, Warren Samuel, 1878-", since it's not constructed according to AACR2 rules. It would only be added as a pre-AACR2 reference with "$$w nna" if it had been established with an online authority record in that form. If we submit an authority for Fisher through NACO, it will omit the "Fisher, Warren Samuel, 1878-" form. I know that doesn't make practical sense, since this is the form most likely to be encountered on older catalog records; but that's the rule we work under. Once the bib headings are updated in MNCAT, our need for the pre-AACR2 reference will drop off, too.
Even if you have a source where the author appears as "Warren Samuel Fisher," LC would be disinclined to add a 400 for "Fisher, Warren Samuel, b. 1878", since the rule is to disregard differences past the first forename except in unusual cases when making references; so this reference would be considered--well, not exactly the same as, but insufficiently different from "100 Fisher, Warren S. (Warren Samuel), b. 1878".