That's how scientific research moves forward.
For centuries, it was considered scientifically proven that heavier objects fall faster. And then Galileo came along, and it was scientifically proven that all objects fall at the same rate. And we still use that simplification today, as g=~9.81m/s2
And then Newton came along, and it was scientifically proven that "falling" is entirely the wrong term: all objects are pulled towards each other with a force proportional to the two objects' masses. And we still use that simplification today, as g = m1 x m2 x G (which, on the Earth's surface, means an acceleration towards the center of the Earth of ~9.81m/s2).
And then Einstein came along...
Science doesn't say "This is what is." Science says "This is what I observe." And then you use those observations to try to predict what you'll observe next.
If you can come up with a model which explains your observations, and predicts, within a reasonable level of measurement error, what is going to happen next, it's a good model. At least, until someone finds a better one.
Science isn't able to say what the fundamental underlying fabric of reality is. However, it is able to say, "This is a model of one possible fundamental underlying fabric of reality, which matches all of our observations to date, and has proven reliable at predicting future observations."
So, no, we can't say if we're all made up of eleven-dimensional vibrating strings, or coalesced energy, or data flowing around some hyperadvanced alien's simulation. But we can say, to use your own example, "The theory of evolution matches our observations about how organisms change over multiple generations, it has proven reliable at predicting future changes of the same, and no other model has proven superior to it, therefore we will accept it as our model of reality until we find a better one."
And that kind of use of the scientific method works whether our perceptions are lying to us, or not. As long as our perceptions are consistently erroneous, and we can model things in such a way that our models work both for past observations and predict future observations, it doesn't matter how well they represent the truth about the underlying reality.
For most purposes, "things fall" is sufficient.
For nearly all other purposes, "things fall with an acceleration of ~9.81m/s2" is more than sufficient.
Neither of those is an objectively accurate representation of the underlying reality of how gravity works. But, unless you're modeling objects with sufficient mass to itself have a significant gravitational pull, the models don't need to be correct. They just need to be useful. And they are. And so is the rest of science, even if our concepts of why it is useful are completely off-base.