Bennett on the Ideological Costs of "Low" Ontologies (Part II: Composition)

In a previous post, I discussed one of the two cases that Karen Bennett focuses on to argue that what one gains in terms of ontological simplicity is (usually?) lost in terms of ideological simplicity. I will now discuss the other case: the debate over composition. In this case, the "low-ontology" side of the dispute is occupied by the mereological nihilist who holds that simples never compose a whole.

Bennett's charge is that the nihilist's "low" ontology comes at the cost of "high" ideology. Here is how she puts her point:

[In order to recapture claims such as 'these paper clips are arranged in a chain', the nihilist] needs to introduce clever techniques that allow him to talk about the very complicated, highly structured ways in which simples can be arranged. On the face of it, however these very complicated predications of simples appear to commit nihilist to the claim that simples collectively instantiate very complicated structured properties. The simples instantiate (((being arranged quarkwise) arranged atomwise) arranged moleculewise) ... At least, the nihilist is committed to the complex structured plural predicates themselves. Here again, the high-ontologist is not committed to any such thing. The believer [who occupies the high-ontology side of this dispute because she believes that there are things whose proper parts are simples] need not countenance either these highly structured plural predicates, nor any properties that answer to them. She does not need to say that the simples themselves directly satisfy any such predicate or instantiate any such property. She can simply say that the simples directly satisfy 'arranged quarkwise'--or whatever the smallest items composed of simples are. Then the quarks satisfy 'arranged atomwise', and so forth up. It is molecules that get arranged into cells. (p.64)

It's not clear to me that Bennett is successful at showing that the ideological price of nihilism is higher than that of "believerism". Bennett concedes that the believer needs predicates such as 'being arranges X-wise' (and possibly the properties that come with them). Her claim, however, is that the nihilist needs the complex structured plural predicates. But her argument for it seems to be based on a premise that it is, say, molecules (not simples!) that get arranged into cells. But, of course, the nihilst would deny this--according to him, there are no molecules, there are only simples arranged moleculewise and simples arranged cellwise and some simples arranged moleculewise are arranged cellwise with other simples arranged moleculewise. So if he wants to say 'These molecules form a cell', he has to say 'These simples that are arranged molculewise (and these simples arranged molculewise and ... and these simples arranged molculewise) are arranged cellwise' but in doing so he does not seem to be using a complex predicate more than someone who is saying 'These children and these children are smart' is (yes 'being a child' is singular and distributive and 'being arranged moleculewise' is neither but Bennett seems to concede that the believer needs plural non-distributive predicates as much as the nihilist).
So, is believerism any cheaper ideologically? What should the believer say of 'These molecules form a cell'? Bennett seems to think that he could just say 'These molecules are arranged cellwise' but for the believer molecules are presumably sums of parts arranged moleculewise, parts which are themselves sums of parts arranged atomwise, etc. So, it's far from clear to me that she is better off ideologically, for the nihilist could just skip all the inbetween levels when she does not need them (after all, pace Bennett, it's ultimately the simples that are arranged atomwise, moleculewise, cellwise, etc.), while the believer would always have to mention that in order for this mereological sum to be a cell, it needs to have parts that are arranged moleculewise, and these parts need to have parts that are arranges atomwise, and these parst need to have parts that are arranged quarkwise, etc.