When you cite any source, you should give enough that the reader can, at least in principle, read that source for themselves. This information is especially important when you cite something as a source for further information, as it is in your example paper.
This paper presents the framework of such a complete phenomenological model outlined by Sayeedvafa (2002) that provides a description of a wide range of the observed behavior, which are both tractable from analytical as well as computational viewpoint.
For example, citations of technical reports (like your first example) should uniquely identify both the institution and the report, so that the reader knows who and how to ask for a copy. In particular, if the report is available on the web at a stable location, the citation should include a stable URL.
If a source is likely to be inaccessible to most readers (like your second example), you should also cite an accessible secondary source that describes the relevant content in detail. (If you really want to be helpful, the primary source citation should include a pointer like "Cited in [xxx].") Otherwise, you're just asking for the reader's blind trust that the source has the missing details you claim, or proves the result that you claim, or is as important as you claim, or even exists at all.