Ok, so it seems that "ternary" is "exhibit A" and the poster child example of why digital is not the same as binary.
The claim is that +5V, 0V, and -5V are discreet levels, ie they are digital and not continuously variable; however it is noted in this example that these 3 levels are not really exact discreet levels but ranges. ("The logic levels are also defined, although obviously the discrete input levels are not precisely defined as shown, just as TTL levels are not precise but given as ranges.").
Presumably in this argument we don't want to call the ranges "analog" because they are not "continuously variable"’; rather they are values within "ranges." Seems like skating on some random ambiguity as many analog signals are not infinitely variable but rather fall within "ranges." Nonetheless, we will proceed on the premise that these 3 ranges are not "analog" but rather they are "digital" values. Fair?
Next, after all of that ^, we proceed to show how these three values (each representing some range) are designated as 1s and 0s (binary digits): 00, 01, and 10. So in the end, we have taken some value that we insist is not "analog" but digital and we represent the value with binary digits.
It seems that in terms of the output, ie the representation of non-binary digital values in this digital system, the example shows the output to be binary. So, if this is our poster child example, we can't argue with the fact that the output of a digital system is binary because this example shows that the output is binary.
All that is left then is to make the case that the input of this digital system is not binary - which is based on the premise that this digital system has 3 possible states (some range in the vicinity of +5V, some range in the vicinity of 0V, and some range in the vicinity of -5V). This then, is the basis for the argument that digital systems are not necessarily binary.
If this is a fair restatement of the discussion, then we can say that the input of a digital system might be something other than binary (ie, something with more than 2 states), but the output (at least in this poster child example) is still binary.
Now, before going too far into the rabbit hole, someone might say "but the output isn't really binary either, it isn’t two states because the values (as shown) are 00, 01, and 00. So according to this logic there are three states. Personally, I would say "tilt" on such an argument because as has been said over and over throughout this thread, the notion of binary is that two states (0 and 1) can be used to code many other values subject only to the number of bits within a "Byte". So, I think we have to agree that the output of a digital system is binary (unless it gets converted to something else, such as analog).
So unless someone wants to argue otherwise (and I’m guessing someone will), it seems that the “ternary” example is consistent with a digital system in which the output is “binary.”
This leaves us with whether the input of a digital system is binary, or if in fact the input of a digital system can have more than 2 states. Again, in our poster child example the input has 3 states, and even though the poster child example clarifies that these 3 states are not limited to 3 discreet values but are in fact values that fall within 3 ranges, we choose to refer to these 3 ranges as "ternary." Whether "ternary" means exactly 3 values (which it doesn't in the poster child example) or whether it means roughly 3 values (which it does in the poster child example), we can all agree that the input has more than 2 values. However, if we accept that these three values are not discreet, but are in fact ranges, then we have to acknowledge that the values could be many values more than 3. For example, approximately +5V could be 5.000000V or 5.0000001V or 4.999999V or pretty much any number of values if we are willing to move the decimal point. Nonetheless, in this poster child example we say it's "ternary" and we use the ternary example to defeat the idea of digital as meaning binary and we insist that because there are (at least) 3 values in the input, it is therefore an example of a "digital" system and therefore "digital" cannot be synonymous with "binary". Or so the argument goes.
If it ended this discussion on a consensus, I'd be inclined to say "ok, if we can accept that the output of a digital system is binary, ie 1s and 0s, ie, binary digits then sure, good enough for discussion sake if not an exact definition." But I have a hunch this thread might have a bunch more posts to go, so what the heck, let's keep going.
So, next, beyond the output side of a digital system being binary, let's revisit the idea that "ternary" is a special case that has 3 states. As we've seen above, the 3 states are not discreet, they are ranges, and the ranges could represent a huge number of actual values depending on where the decimal point resides. But let's give the poster child example the benefit of the doubt; let's call it 3 states and say "ah ha, it is 3 values and 3 is greater than (and definitely different than) the 2 states of binary – so therefore digital can’t mean just binary.” This then raises the question of "who cares"? Because darn near everything we input to a digital system (all the alphas, all the numerics, all the special characters, all the pixels, and pretty much all the everything) has more than 2 states. But you know what? When we get done coding all the darn near everything (including in the poster child ternary example that we insist is "not binary") what the digital system does is it converts the 3 states (or n states), or ranges (with an infinite number of possible values) to 1s and 0s, ie to 2 states we call "binary." Again, how many 1s and 0s we use to encode something else is just a function of how many values we want to represent with binary digits as we string the bits together into Bytes according to various coding conventions.
If we zoom out and look at the huge amount of information inputted, processed, and outputted with computers and embedded devices and all the information collected and distributed with networks the vast vast vast majority of it gets reduced to 1s and 0s. The argument that digital and binary are not synonymous might be more academically interesting than useful.
(edit for typos and clarification)