Lab simulations are reality...however, they are necessarily simplified. Take bacterial evolution experiments. For the most part they involve testing for mutations involving some kind of resistance or tolerance i.e. antibiotics, heat, substrate utilization. All other variables are controlled and do not vary (or don't vary much). Also, the amount of genetic variation in the starting population is often controlled so that you know what you started with so that down the line you can observe what has changed directly. This allows the outcomes to occur more rapidly because the evolution is directed in a sense i.e. you are selecting for resistance etc. and seeing how the bacteria get there. In nature, there is not a homogeneous gene pool starting point and different selective pressures are acting on the population simultaneously such that overall the gene pool will go towards whatever is the best consensus for its environment..if the environment changes, it may completely shift in another direction. Though people do study such natural populations, it is a hell of a lot harder because you have many more variables. That does not mean that the lab experiments tell us nothing about what happens in nature.
Think of it like Mendels pea experiments. He picked traits that demonstrated the principles of single gene inheritance patterns. They are hardly representative since most phenotypes are quantiative and controlled by multiple genes..but again, that is much harder to measure than the traits mendel picked. He studiously avoided the harder traits. However, his simplified "lab" experiments yielded results that are still relevant for quantitative genetics.
™ Version 4.2