Hi Josh,
You want to use either a flexible factorial model in SPM (with a subject factor) or one of the GLM_Flex models (I'm not sure which version is best, you can ask that mailing list for the current model to use). Alternatively, if you are not interested in the 3-ways interaction, that is unlikely to significant with 3 subjects, you could collapse some one the factors and use the SPM flexible factor out of the box with the subject factor.
I believe there is a way around SPM's limit of 3 factors and that you can use 3 factors plus the subject factor. There has been some recent posts around this issue.