Latour's modes do not describe their referent practices as they are actually practiced. The modes describe their referent practices as they are practiced well. In the case of religion that might mean that the religious mode has been more or less dead for centuries. In the case of politics, well I wonder whether it has ever actually worked like that. In the imagination of 'establishment liberal' John Dewey, perhaps, but nowhere else.
Latour's 'values derived from experience' are not, therefore, value neutral. On the contrary, they derive from a definite value judgement in the normal, non-ironic sense of that term. These values are inseparable from their author, at least initially. He may try to pass his experience on to us - and this is right, this is good, why else would we read his books? - but we must not allow him to step away from his instaurations so easily. He must be made to work and sweat to give them their own lives, their independent trajectories. So far his achievements in that regard have been mixed.
The modes are most convincing when they are actually based on fieldwork. The reference mode is an updated version of the famous chapter from Pandora's Hope on circulating reference. I think that is still the best essay he's ever written and it'll probably be the one that is still read many years from now. The technology mode is also if not entirely convincing then certainly compelling, especially if you've read his book Aramis. The law mode isn't particularly well explained in aime but it makes sense in light of his book The Making of Law. The problem that mode has is the sheer plurality of ways that law is practiced. However, again, because it is actually derived from fieldwork it makes a strong case (so to speak!).
Where Latour goes awry is in the chapters that aren't based upon fieldwork but upon other philosophers. The politics mode is, as I suggested above, lifted straight from John Dewey; in particular The Public and Its Problems. To the best of my knowledge Latour has never done any empirical research on politics whatsoever. His overlong and, considered alongside the rest of his corpus, rather superfluous Politics of Nature is entirely unblemished by any kind of experiential concreteness. Dewey also seems to provide the basis of the religion mode with his A Common Faith.
Latour has many more philosophical influences besides Dewey, of course - Lippmann and James just to name the other early twentieth century Americans. The reproduction chapter is a bit like reader's digest version of Whitehead; the whole 'Gaia' and 'end of Nature' thing is derived from Michel Serres' The Natural Contract; most of the concepts that date back to Irreductions are Nietzschean through and through. And so on.
Of course, it isn't that doing philosophy is bad. In fact Latour is a terrific bricoleur of philosophy. One of the best. The problem with these 'pure philosophy' modes is that they take the abstraction so very far away from practices but nevertheless maintain the same rhetorical insinuations, continually implying themselves to be concrete, hard-nosed and evidentialist. These chapters radically conflate the strictly philosophical meaning of 'empiricism' and the other sense that indicates fieldwork and evidence. They might be empiricist inasmuch as they operate on the basis of a concept of experience derived from James' radical empiricism but they have little to no empirical content in terms of fieldwork, case studies or even anecdotal examples.
These purely philosophical modes are much less convincing. They cling to their author; he is the only thing holding them up. The politics mode, for instance, is scarcely deserving of the name. It is a theory of democratic politics, for a start; it should be called [dem]. It does nothing to describe politics as it is practiced, in fact it makes that more difficult. It is a work of political philosophy that attempts to describe the processes through which the liberal democratic ideal could be realised. It operates in a space of pure abstraction.
There's nothing wrong with that as such but it is squarely a matter of political philosophy not political science. It is, therefore, more or less useless for any actual fieldwork. In fact it is worse than useless because, operationalised, it would only distract fieldworkers from what is in front of them - the phenomena they should really be paying attention to. In other words, it would sunder the most fundamental tenet of actor-network theory (and ethnomethodology), to follow the actors and learn from them.