Does anyone ever proved that by introducing new compatible axioms we always narrow down the set of possible valid deductions? I think this is intuitively correct (but the fact it is false may in fact be counterintuitive).
So given some axioms we prove theorem $\Bbb P$.
Adding further axioms can't increase the number of cases where we can apply the theorem. So if $\Bbb P$ is our result, adding new axioms to our "language" just reduce the result to something smaller.
If I understand correctly most times people try in fact to remove as much axioms as possible from the "language" and also to remove as much hypothesis as possible from theorems to make them applicable in most cases.
You need to distinguish between axioms and assumptions of a theorem and between one theorem and the body of theorems derived from a set of axioms. A set of axioms, like the real numbers, the group axioms or Peano arithmetic, is used to prove many theorems. Some of those theorems have the form of implications. For example there is a theorem in analysis that if the function $f$ is continuous on a closed interval, it attains its supremum and infimum. You could strengthen the assumptions by asking that $f$ be differentiable. That weakens the theorem because it applies in fewer cases than the earlier version. If you add to the axioms, you increase the set of theorems you can prove because you have more at your disposal.