The first mistake is not claiming at all, usually because the work feels too clinical to count. The second is under-capturing eligible salaries and overhead. Those costs are often larger than expected once you account for the time actually spent resolving uncertainty.
The third is a technical narrative written for peers rather than for a CRA reviewer. The science may be sound, but if the document does not frame the uncertainty and the experimental approach in the program's language, it invites questions.
The fourth is poor contemporaneous records. The fifth is filing late. Both are fixable with a process, which is precisely what a specialist is for.