My most critical assumption with cellulosic biofuels is on land efficiency: tons of biomass per acre, and hence gallons of fuel produced per acre, and more accurately, miles driven per acre. I believe biomass yields per acre will multiply by two to four times from today's norms.
The lack of genetic optimization and research on cultural practices, harvesting, storage, and transport with would-be energy crops -- miscanthus, sorghum, switchgrass, and others -- means that there is significant potential for improvement. The application of advanced breeding methods like genetic engineering and marker-assisted breeding, limiting water usage through drought resistant crops, and large-scale application of biotechnology (i.e., optimizing the process by which plants conduct photosynthesis, or reducing stress-based yield losses) will also contribute to increased yields with fewer inputs.
More importantly, different energy crops are likely to be optimal for different climates -- jatropha makes sense on degraded Indian land, but not in the American Midwest. Rather than a single dominant energy crop, we are likely to see a variety of feedstocks that allow specialization to local conditions, mixes, and needs, while mitigating the risks.