Look, here is the story:
1. You will not find any meaningful data on this, anywhere.
2. The "gold standard" for physician salary reporting in general is the annual MGMA Productivity and Compensation survey; you should be able to find it in your friendly neighborhood med school library. It gives pretty good data on neurology (and other specialist) salaries in general.
3. Fellowship training only gets you more money if the following three conditions are met:
a. during fellowship you pick up procedural skills that bill/reimburse at a high rate (for neurologists this means EMG, sleep, botox, +/-EEG), and
b. your job allows you to perform lots of these, and
c. your compensation is determined by how much you bill/collect, as opposed to flat salary.
Now, there are lots of interesting subspecialties in neuro, but many (neurobehavior, MS, movement, headache, to name a few) are really still just history-and-exam based, meaning you spend a lot of time talking to and examing the patient and writing some Rx's, all of which is very time consuming and on an hour-by-hour basis doesn't get you much financial return. In the time it takes to do one really good comprehensive neurobehavior/dementia exam, you can have done 3 EMGs or popped botox into about 10 patients.
You are correct that you don't "need" a fellowship to do/get paid for many neurology procedures. If you had a solid residency program, you can do lots of stuff reasonably competently. So you are correct that in many cases a fellowship is primarily an exercise in "skill sharpening"
That being said, the learning curve between residency and fellowship training for most "procedural" stuff in neurology is pretty steep. Having done fellowship training, I can tell you with confidence that I am, oh, about a million times better at neurophys test interpretation than I was prior to fellowship, and I had pretty good residency training. Now, you might say, "so what, you can pick that up on the job as you work." That may or may not be true, and is highly dependent on your work environment. If you're working at Big Sucker Academic Medical Mecca with dozens of subspecialists running around giving lectures and showing you how to do stuff to really esoteric patients, you may well pick up a lot. But the vast majority of practicing neurologists are in solo or small group practices where they don't have a lot of academic support and are doing a limited amount of mostly bread and butter cases: lots of carpal tunnel EMGs, lots of normal EEGs on patients with something other than epilepsy. You don't really get better that way. Who ultimately suffers for your relative lack of experience? That's right, it's the patient you end up misdiagnosing or sending out for a zillion other unnecessary tests.
So, if you plan to be an office based, low acuity general neurologist, and aren't gunning for big bucks, I'd say you can safely skip a fellowship. But if you a) have a burning desire to subspecialize, b) want to be really good at cranking out highly reimbursing procedures, or c) just want to "be all that you can be," then you should probably do a fellowship.