Dear Andrew,

I think Nathaniel's post is similar to the general issue of also continuously evolving and changing annotation platforms and annotating generally microarrays through various annotation packages-- there is an example with affycore tools below (via select, etc):

https://www.bioconductor.org/packages/release/bioc/vignettes/affycoretools/inst/doc/RefactoredAffycoretools.pdf

In detail, it is mentioned also in the paper: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1283542/

One is the probe set redundancy, which is the case you mention above--but also there is the other case:

"Non-specific probes"

A significant increase of cDNA/EST/genome sequence information leads to the possibility a probe thought to be specific for one gene can actually hybridize to transcripts from additional genes or non-coding transcripts. As shown in Table 1, according to the current version of the UniGene database, for most GeneChips 10 - 30% of probe sets contain at least one non-specific probe. Probe alignment to genomic sequences also reveals that 5 - 16% of probe sets contain a probe(s) with more than one genomic sequence hit(s). The difference between the UniGene- and genome-based criteria may largely be due to UniGene clustering or EST sequencing errors.

Best,
Efstathios

Unfortunately(for some reason I didn't get any updates for the new comments and answers) I believe Nathaniel you have probably moved along with this issue---just to pinpoint my opinion on this matter: there isn't a clear option on this specific issue, but you can move similarly to the above vignette in affycoretools and with the most "naive way" keep the first mapping like:

https://www.bioconductor.org/packages/release/bioc/vignettes/affycoretools/inst/doc/RefactoredAffycoretools.pdf

Alternatively, you could consider the above approach i have mentioned, and download and proceed with customCDF arrays--

Hope that helps,

Efstathios