It sounds like you understand the details of getting data from GEO and taking .CEL files to an ExpressionSet. Where things are going to get complicated is in getting "all of this information into a single data matrix for analysis". Doing so may not be the best approach, but it is impossible to know without a good deal more background, a level of background that is not easily communicated in a forum or email. Since you say you are relatively new to the whole area, I suggest you find a local bioinformatics collaborator who can work through the data with you.

If it was me, I would try hard to get CEL files for all of them. If not in GEO you might try requesting directly from the author. I've had about 50% success with this in the past. Then with all CEL files, you can use affy, gcrma, and custom cdf to create one consistently summarized and normalized dataset mapped to gene symbols (Retrieving Probe To Gene Ids For Affymetrix Chips In Bioconductor). The latter would allow you to compare with any datasets at gene level where you don't have raw CEL files. Whether you are able to process all together or you try to combine differently processed data after the fact I would be VERY aware of the potential for batch effects.