EM: Consider a model of the business, say implemented in Excel. This model would project all cash-flows and for each year-end reserves and required capital can be calculated. This model would have some parameters that are based on assumptions. Assumptions can always be refined by doing more thorough investigations. But which assumptions warrant further investigations? Surely the ones that has a large influence on the final result. For example, suppose we change the equity risk premium assumption from 4% to 6% and find that it does not really affect our results, then we would not bother to research the equity risk premium to find a better value. If we find that changing our assumed long-term inflation from 5% to 6% has a large influence on results, then we would definitely put some time into getting a good value for assumed inflation. This kind of analysis is called sensitivity analysis: getting to grips with which assumptions really affect our results, and understanding the size of the effect.
EM: With term assurance the size of the reserve is reasonably small, most risk premiums are paid out in each year, subject to mortality randomness. Due to the smallish size of the reserve, investment rates do not play a big role. As you suggest, mortality rates would be significant.