This is going to look horribly messy while I hunt down references and stuff to put in. Please feel free to not use the Discuss tab.
The gismu list contains portions that are enclosed in parentheses or square brackets. It is a contentious issue whether these parenthetical remarks have any normative bearing on the meaning of the gismu, and if they have, what kind of interpretation they should have.
This problem is exacerbated by the fact that an effort is underway to define as many lujvo as possible with definitions in the style of the gismu list. These definitions must now be called into question, to the extent that the authors' understanding of the meaning of parenthetical remarks differ from that of the gismu list, and from each others'.
See also: BPFK gismu Section: Problems With ka.
It is sometimes said (e.g. in Cowan 1997) that gismu are intended to "blanket semantic space". This has led to the emergence of the view that the meaning of gismu should be as wide as possible. It is easy to narrow down a sense by constructing lujvo or fu'ivla, but it is much more difficult to combine gismu to construct a sense that is broader. Hence, if there are two competing interpretive conventions for gismu definitions, one which consistently results in narrow senses, and another which consistently results in comparatively broad senses, according to this view one should prefer the broadness-maximising convention.
On the other hand, Cowan (1991) also notes that gismu should not be seen as semantic primitives.
John Cowan, 1991. Loglan and Lojban: A Linguist's Questions And An Amateur's Answers. http://www.lojban.org/files/why-lojban/reply.txt
John Cowan, 1997. The Complete Lojban Language. Logical Language Group
Old gismu discussion from 1994: http://balance.wiw.org/~jkominek/lojban/9409/msg00111.html