When you place a sample in between the photon source and the detector, you want there to be enough of the sample that many of its nuclei interact with the gamma rays emitted by the Co 57. However, you don't want your sample to be so thick that few of the photons actually make it through to the detector. In extreme cases, the Mossbauer spectra may even suffer from thickness-related distortions. In other words, minimize Compton scattering and maximize resonant absorption. The optimal thickness for your samples will attenuate the count rate from what it was with no sample present by a factor of e or about 63% (see online book, section 'Absorber Optimization' - http://link.springer.com/content/pdf/10.1007%2F978-3-540-88428-6.pdf).
Of course, we can't adjust the thickness of our Fe foil (it's approx. 1 mil), but we can vary the amount of Fe2O3 and Fe3O4 powders we use. It is likely that the samples you will be using are small pieces of aluminum with half-inch punch-outs (area: 1.27 cm^2) containing the Fe2O3 or Fe3O4 powder. Try to make sure that the powder is approximately evenly distributed within the punchout. A note about accurate sample alignment: The area of the face of the silicon detector is quite small, but the samples barely cover it. Be sure to line the samples up very closely with the center of the detector face, since even a few millimeters too high/low/to the side will allow more photons to enter the detector without even travelling through the sample. The result of mis-alignment will be a spectrum with a poor signal-to-noise ratio, which will be difficult or impossible to analyze! Be careful to limit your exposure to the source when lining up the sample with the detector. You might want to (carefully!) remove the detector from the setup and use a piece of tape to attach the sample directly to the brass collar of the detector to get the best alignment possible.
It works best to err on the side of fewer counts, rather than more. Sure, you might be able to accumulate lots of samples quickly, but you are likely to have a poor signal-to-noise ratio. If you have relatively few counts, you are more likely to have positioned the sample well and to see an interaction - it's absorbing the incoming gamma rays as expected. I suggest using at least 71 mg/cm^2 of Fe2O3 (or about 90 mg total in a 1.26 cm^2 cutout), and at least 48 mg/cm^2 of Fe3O4 (or about 60 mg total in a 1.26 cm^2 cutout).
Next, Data and Calculations