I think that in the end, all the inconsistency works out as seamless consistency...
In most of the incarnations of Star Trek, shields have explicitly been multilayer affairs: the computer graphics aboard the starships show multiple shield layers, and occasionally even the combat VFX makes it clear that the shields consist of nested shells. An example from "Datalore":
Now, we could easily argue that we are simply witnessing different levels of penetration. A weak opponent will only manage to light up the outermost bubble shell(s), but a strong one will penetrate all the way down to the skinmost fields.
On the defending side, a starship's outermost shields will take the punishment if they are strong and modern by design and
the ship has a capacious power plant to energize them and
is allocating power primarily to shields. In combat of the sort we see in DS9, power is probably allocated elsewhere, and the opponents are fairly strong in relative terms; the screencap above is of a rather feeble opponent confronting a modern and large ship that has no other power allocation needs...