Fee isn't limited to what you're paying the exchange, you have to take into account slippage - the price you use in your backtests is NOT the price you could actually realistically buy at because as soon as you buy, you move the price (especially in size).
Strictly speaking, you should be incorporating level 2 order book data to properly gauge fees and simulate real world profits from your backtests.
You can (and I do as well) - although some proprietary bots have had notable problems with figuring them out.