How do you handle backtesting models when spread and commission data are missing from past charts?

Backtesting software often ignores spread widening during session opens. How many pips do you manually add as a penalty buffer during backtesting to keep metrics realistic?