I don’t know everything about all your questions, but I think I know a few things, so I’ll share them.
At the sentence level, consider a string of 平板 words; that is, the entire sentence is flat. In that case, there is a gradual pitch drop over time. This is useful to know because when you are trying to listen for pitch, your goal shouldn’t be to ascertain if pitch lowers, but rather whether there was a significant enough drop for it to be intended as a downstep.
Another observation at the sentence level is something called “terracing.” It’s basically the idea that each subsequent pitch drop within a sentence is going to be less and less dramatic. This combines nicely with the idea of gradual pitch drop over time, because then every time you raise your pitch to drop it (eventually), that pitch is going to be a little bit lower than what it was at the start.
The gradual pitch drop and terracing effect are “reset” every time there is a pause, usually because the sentence ended, or because the clause ended (e.g.: 日本人ですが、アメリカ人らしい - by the time you reach が, your pitch should be low, but then you start a new clause and you start from scratch at your baseline initial high pitch again.)
A specific fact that’s good to know is that two 平板 words put together just stay flat (with the gradual pitch decrease I mentioned before - by the way, this happens naturally; it’s very hard to stay at the same pitch the whole time). In other words, although the graphs may show the individual 平板 words as starting low and then going and staying high, I never ever think about the first mora as low for a 平板 word because it’s almost impossible to start at the exact same pitch anyway. Besides, in Japanese, a rise in pitch is only important insofar as it helps you drop your pitch on some future mora. (The only exception being questions, I’ll get to that later.)
As for informal questions without the use of か, this is actually a place where Westerners tend to make a lot of mistakes. The pitch family of the word stays the same! But yes, there is a slight rise in pitch at the end when it’s a question. But the key takeaway is that you should still pronounce the word in the correct pitch accent family, just with a slight rise in pitch as you are finishing pronouncing the last mora of the word. So if you are asking, “is it a dog?,” don’t suddenly change it to い↓ぬ？. Instead, it’s something like いぬ↑？ with there never being a downstep (or else it would change the pitch family of the word)
).
