When would you prefer masked contrast images over traditional SPM t-maps?

It depends on what you want. If you are interested to locate a biologic effect as precise as possible, I'd prefer masked contrast images since they are independent of the amount of local noise. Local noise comprises many components from measurement error to biologic variance deriving from other neurons that can be found in the same brain region but serve other purposes. Why divide by those if you are not interested in them?

On the other side, you may be interested in regions to be used in a second analysis. In a PET study, e.g., you may search for brain regions that differ between patients and controls and use these regions to create a diagnostic index in further patients. Or you may identify brain regions with a significant group difference (e.g. patients versus controls) and extract data from these regions for correlation analysis in one group (e.g. disease severity, disease duration etc.). In that case you probably prefer the original t-map clusters, since you are interested in a region with a good signal-to-noise ratio.

At least, you should at least try out mascoi, so you get a feeling for what you have been missing so far ;-)

I am getting the error message
sorry, only traditional voxel sizes (multiples of .5mm) are supported

I'm sorry for this. I started mascoi by copying and pasting some matlab routines I've been using for quite some time. The algorithm by which the sections are created is one of them. It is not incredibly flexible, but I like the image quality it produces. I know, I could use one of SPM's routines to become resolution independent, but I do not have very much time for coding at the moment.

What can you do if you get this message? You can still use mascoi for its basic functionality, i.e., obtain a list of suprathreshold clusters, see the stats that describe how "masked contrast images" differ from the original clusters, select a subset that will be displayed in colored maximum intensity projections and export the 3D masked contrast volume. However, to generate sections, you'll have to browse the created volume with an external viewer.