Inaccurately assessing cashflows. This is the single largest mistake. Generally, you know the costs. Where people err on on estimating the offsetting revenue. Sales projections tend to be overly optimistic. To account for this, include three separate sets of cashflows: optimistic, most likely, and pessimistic.
 

Presenting data in the report. For a report to be business intelligence, it can't be a collection of nicely-formatted data. It needs to present the conclusions based on the data. NPV is one of those conclusions. You do not need to chart out the cashflows. You need to show what they mean.
 

Including data in the report. By this, we mean actually including the data instead of linking to the database so when the report is refreshed it's working with current data. This is a huge mistake that creates all sorts of problems, including information silos and conflicting reports. If a report exists for archival reasons, then data capture is a feature of that report. But if the report exists as business intelligence, it must use the central data source instead of becoming a data liability.
 

Not having rules to keep the source data limited. For example, an engineer at a manufacturing plant consistently won approval on his capital projects. The NPV for these was always impressive. He included his valuations of things such as safety--which he wasn't qualified to assign a number to. Consequently, his capital requests were assessed on overstated cashflows instead of on their merits. The problem was later fixed by limiting cashflow information to exclude such things as good will, safety, and employee morale.
 

Failing to validate the source data. Data validation is critical to accurate reports. Implement this as much as is practical. At the very least, implement some basic validation rules in your forms. But also talk with senior managers and department managers about processes and policies that can help prevent the entry of bogus data. It all comes down to people, and you can't automate their ethics.
 

Failing to test the report. NPV is a complicated calculation. In the days when it had to be done by hand, it was seldom done. There really is no practical way to check it manually, so we have to trust the software to do it. So far, so good. You just constructing a report that pulls cashflows from a database and applies an NPV calculation. But when you think about this, it becomes apparent that several problems can occur along the way. That means you could get an incorrect NPV. Test the report by plugging in test numbers. Be sure to include null values, negative values, and values that are deliberately outsized. See what happens.