This is mammalian tissue we are talking about, right? Unfortunately, I do not have much experience with vertebrate tissues in general. As the section surface is rather small in the end, it might be better to cut the fresh samples to appropriate dimensions directly befor fixation, e.g. cutting them submersed in chilled fixative (minimizes crucial time between excision and fixation), to get a better and faster penetration of agents. With my samples, I also had good experiences in general from replacing cacodylate with HEPES buffer, and using a formaldehyde/glutardialdehyde mixture (diluted Karnovsky´s solution) as first fixative, followed by osmium tetroxide diluted in buffer. I´d check the references for appropriate pH values and osmolarity of the fixative. Otherwise, TEM protocols for soft tissue should generally work, but some details depend on the circumstances of your study (choice of resin, dehydrating agent, sample processing, etc..)
time of tissue getting and fixation is important (after obtaining the tissues -1to 4 mm-you have to put them in chilled 4 C glutaraldehyde immediately)
Try to avoid Osmium Tetroxide (OsO4) because it kills most antigens. For contrast just use Uranyl Acetate and Lead Citrate. If you have LEICA freeze substitution machine, it is just perfect.
If antibodies are to be used for markers, I´d also try to avoid the glutardialdehyde and maybe stick to formaldehyde (which also preserves some fatty acids, compensating the missing OsO4). If you use GA with immunolabelling, better stay below 0.5% and below 15` fixation time. Replacing GA by picric acid in combination with formaldehyde is another possibility. Best structural preservation I know is achieved by high pressure liquid propane jet freezing, followed by freeze substitution including the formaldehyde, but not everyone has the necessary equipment (we don´t...), and you are restricted to 200µm sample thickness. Protocols following Tokuyasu are an other way to get good immunogold labelling on cryosections for TEM. As buffers, PIPES and HEPES are preferred over PBS or cacodylate if immunolabelling is to be achieved.
If you plan to do "only" conventional TEM for structural analysis and don´t need antigenicity, diluted Karnovsky works fine. I usually fixate on crushed ice under light vacuum (water jet pump).