hey! So i'm not sure exactly what the context of the water uphill or the two headed horse metaphor was (feel free to provide it if you'd like to discuss it specifically) but I may be able to speak a bit about science in general that may help you:
Science provides a model of reality. A model is an abstraction that can be used to form hypotheses and to run experiments. A model is really about simplifying and removing variables so that the relivant parts (the ones you are interested in) are isolated; remove too many variables and your model is meaningless (it coarse a representation), leave too many in and your model may be too complex (you may not be able to clearly see a pattern or predict results).
So by definition a model is always wrong...it is not reality it is an abstraction. Science is a vast collection of these models (theories, laws, hypotheses) that in aggregate make a very good representation of the reality we live in (and some that we don't live in). The work that is done in science is always disproving past theories and replacing them with models that are a better fit to reality. If science were correct (in the absolute sense) there would be no more need for scientists because we would have a perfect model of reality that we could rely on to give us correct predictions of all future possibilities. this of course will never be the case.
The cool part about science then is that we do not need to believe any of it at all. Scientists have a method that is used over and over to disprove and create better models. so it is not the models themselves that is the science but it is the method that scientists use to reach conclusions about reality that power their models (equations, hypotheses, theories, etc).
So bringing it back to your point, depending on the model you are using, you can both have consistant predictable behavior, and exceptions to that behavior. Einstein's theories of relativity did not help much in explaining gravity that we experience all the time consistantly but they did explain the exceptions. An engineer working out details of a new car engine that will never travel in deep space or have to use the gravity of a star or planet to change speed or direction will certainly not care about Einstein's theories in this context and use equations which are a more crude model of reality. So Einstein's theories approximate reality much more exactly and thus enabled science to move forward in a very real and meaningful way, and did disprove the "laws" of gravity as they were understood at the time, BUT Einstein's theory is just as "wrong" as Newton's it is just the best thing we've got so far. So for the situation of the engineer building the car, using Newton's wrong theory is ok, because what he is trying to predict is within the assumption's of Newton's model. If the car was sent into deep space it would certainly be easy to see the inconsistancies, but here on earth it provides a clear predictable consistant behavior.
In essence, Science at it's very core is about exceptions to consistant and predictable behavior.