Martin (1992: 176-7):
A second factor underlying the differences in categorisation has to do with the essential indeterminacy of some of the relations themselves. The problem of alternation (or) and contrast (whereas) is a case in point. Both relations imply a system of terms (a taxonomy of some kind) — for example, tea or coffee as hot after-dinner drinks.
The terms themselves can be viewed in two ways. One way is to view them as available alternatives, which combine in text like additives:
A OR B : A AND B ::We have tea or coffee : we have tea and coffee ::Get tea. — Or coffee? : Get tea — And coffee? ::They drink either tea or coffee : They drink both tea and coffee ::Have tea, or coffee instead : Have tea, and coffee as well
Another way of looking at the terms is to consider how they are similar (as a result of subclassifying after-dinner drinks) and how they differ (since they are oppositions in the same system). Looked at in this way not A but B contrasts with A similarly B:
A WHEREAS B : A LIKE B ::Tea doesn't taste bitter to me whereas coffee does :Tea tastes just as bitter to me as coffee does ::
I take tea with milk but coffee without :I take tea with milk as I do with coffee ::
So you have tea in the morning while she takes coffee :So you have tea in the morning just as she does
Since the relationship between such terms can be looked at in different ways, classifying the relevant hypotactic conjunctions becomes problematic (e.g. while, whereas, apart from, without, except that, instead of, rather than, other than). Halliday (1985) groups them with additives under the heading extension, emphasising the relationship with alternation ('or') and addition ('and'); Martin (1983) on the other hand focusses on on the idea of opposition ('whereas'), grouping them with similarity ('like') under the general heading of comparison. Because of the indeterminacy of the relations themselves, neither categorisation is completely satisfactory.
This is quoted at length because it is presented as an argument that is intended to identify problems in the categorisation of logico-semantic relations, as part of the justification for the 'discourse semantic' approach that is about to be undertaken.
 This is not the same relationship "looked at in different ways". The first set of examples display the extension categories of addition: additive: positive (logically: A and B) and alternation (logically: A or B).
However, the second set of examples, instead of focusing on additive addition and alternation, displays the extension category of addition: adversative (logically: A and conversely B) and the enhancing category of means: comparison (logically: A is like B).
In addition to not being an alternative view of the same logical relations, it is manifestly a simple category error to group A and conversely B (adversative) with A is like B (comparison).
 Because of the category error and invalid reasoning involved, the argument fails to make the case for the claimed "indeterminacy of some of the relations themselves".