Your still ignoring on screen dialog about the Defiant, they may not have ever been produced on a mass scale however she was never intended to be a one off
She was intended to be a prototype. Even the prototype versions of things are intended to be prototypes for SOMETHING. But there's no such thing as a "mass produced prototype," as that is a contradiction in terms.
Hence Defiant wasn't supposed to be the first of the new class of starship, it was supposed to be the ship that
helped them decide whether to build more of them. That's an important distinction to make: the only way they would ever build more of them in the future is if and when the first one was proven technically sound. If it wasn't, they would either never build another one, or they would redesign the whole thing and build a new ship with a new design, test that, and if that worked better, build more of them. That is exactly what is meant by "prototype." You don't actually know if you'll build more of them or what the "more of them" will even look like until you've tested the prototype and analyzed the data. Often enough, the prototype has certain problems that can only be fixed with a complete redesign, and therefore becomes a one-off.
The other ships we can't assume just because they were a prototype they were intended to be a one off
Prototypes ARE intended to be one-offs. That is to say, you only build
one prototype for testing purposes. It's only once the prototype proves the design is sound and the technologies can be integrated properly that they decide to build more of them. The thing is, what happens in the REAL WORLD is that the prototype always has certain issues or omits certain technologies that aren't needed for critical tests, and is therefore noticeably different from the production model anyway. Again, it's a lot like how the original YF-22 prototype didn't have a weapon system or helmet-mounted fire control system installed, didn't have full battlefield network capability, and wasn't actually combat ready in and of itself. If they had built the entire production line to the exact specifications of the prototype, they wouldn't have half the capabilities they were supposed to have (of course, they'd probably enter service a hell of a lot sooner, but still...)
USS Midway is another good example of this. It was supposed to be the first U.S. aircraft carrier to have an armored flight deck, but the design performed so poorly that they basically abandoned it and started over with a totally new hull design and never even attempted to build another one. If Midway's design had worked better, it probably would have entered mass production itself instead of the Forestall class.
We have also seen many more original design Excelsiors than Enterprise-B variants through TNG and DS9 which would say to me that there are more of that type than the variant.
Sure, but we have no idea what their specifications are. Judging by the cosmetic differences in the model between TSFS and TUC, it's clear that even Excelsior had to be extensively refit when it finally entered service, so there were a lot of things in its design that either didn't work as intended or were improved after testing. If any of those things had been deal breakers for the design, Excelsior would have been the only ship of its class and Starfleet would have moved on.
That's what it means to build a
prototype. You don't know how well it works until you build it and test it. If anything goes wrong with it, you'll never build more. Starfleet seems to be really good at improving on flawed designs without radically changing them, which explains the small number of single-ship classes. But it is a strong enough possibility that we can comfortably say that some designs probably WEREN'T continued past a handful of test ships, and both the NX-class and the Franklin class could be good examples of this. NX-02's design is sufficiently different from the 01 as to be almost a different class of starship already; I'd bet even money that NX-03 was completed in what would have been NX-01's 5th season refit.