It isn't true either at least not of scientists in general.
Politicians and economists are far more worthy of that saying.
BTW A theorem can be proved from some given set of starting axioms.
Those axioms may or may not be correct in reality but if the chain of reasoning is correctly followed then a theorem can be proved from your chosen starting axioms using formal mathematical logic. Comparison of a models predictions with the real world is the ultimate arbiter.
Requiring the internal angles of a triangle to always add up to 180 degrees give you flat Euclidean geometry. Relaxing that restriction allows what turns out to be a useful generalisation of spacetimes.
I am yet to be convinced that string theory will ever be a useful replacement formalism but only time will tell. Some very bright guys are convinced that it is *the* way forward.
Astrophysicists are painfully aware of the limitations of trying to work out how a forest works by studying only the largest and nearest trees.
It is a sobering thought that 10% of everything in the peer reviewed literature will subsequently be shown to be incorrect or misleading. Even papers that are mostly correct occasionaly suffer proof editors "improvements" that can in the worst case invert the intended meaning of the authors. Letters to the editor can be quite feisty if it happens.