To receive a US patent, the "invention" must pass three tests: 1) usefulness, 2) novelty and 3) non-obviousness. The first is easy to pass, the 2nd can take more effort to determine (and the determination is always subject to later reviews) but is doable while the last is the most difficult due to this question: When is an invention obvious?" Attempts to define obviousness have been going on for over 200 years and are not settled.
The SCOTUS recently came down with a landmark decision in the area KSR v. Teleflex which concluded that if you take two or more mechanical parts and combine them in such a way that they do no more than the sum of the parts, the invention is obvious. i.e., 1 + 1 = 2 is obvious, 1 + 1 = 3 is not, ditto for 1 + 1 = 0. In the second (and third) case, you found an unexpected result. Prior to this, the standard for obviousness was that an invention was obvious ONLY if there was literature stating that you should combine 1 + 1 to get 2. If no one had said 1 + 1 = 2, it was considered non-obvious.
KSR however only applies to mechanical devices. A new appeals court decision Eisai v. Dr. Reddy's Laboratory and Teva Pharmaceuticals has begun to address this issue, and I'm not sure that I agree with the outcome. The case involved a pharmaceutical.
But first some background. With patents on pharmaceuticals and in general, chemicals, the "heart" of the invented molecule is clearly and usually narrowly defined, but is the patents will always claim one or more moieties, "sidearms", for which the exact chemical composition is not critical. The moities only need to have certain characteristics such as water solubility, inertness,... After reading just a few patents, any good chemist would get the idea and be able to see what is going on with the moieties. In fact, these moieties are such a popular idea that they have their own name: "Markush claims" named after the first case that included them nearly 100 years ago. It is commonly understood that the inventor does not actually prepare all of the examples, but just uses his knowledge and experience to claim that they would work.
In this case, molecule with a trifluorethoxy moiety was known in the published literature to work. Eisai filed a patent with Markush claims that dropped the terminal trifluorocarbon and replaced with with (amongst others) a monofluorcarbon. The stated opinion of the judges was "To the extent an art is unpredictable, as the chemical arts often are, KSR’s focus on these "identified, predictable solutions" may present a difficult hurdle because potential solutions are less likely to be genuinely predictable.”
So here we have it both ways: broad Markush claims are predictable, but narrow examples nearly similar to them are not predictible. I'd even be willing to bet that there are multiple patents were the Markush claims would list both the mono- and trifluorocarbons.
Markush claims have been getting broader over the years, now they will really spread their wings and soar.