Heritability is defined as the proportion of variation between two people that is explained by their genetic differences. (Remember that we each possess 23,000 genes, and over half of these come in a different flavor from your neighbor's version of the gene.)
If the variation in height between identical twins (who have identical genes) is 2 inches, this is 0% heritable, because it's not based on genetic differences (there are none). It must be explained by differences in their environment (food, disease, etc).
Yet this idea breaks down, when the environment (nurture) is inherent (and anticipated) in the design of the organism (nature) itself, especially when the nurture acts as a trigger, for which the organism is designed to respond. (See a simple example here).
Also, genes are like algorithms. They accept complex inputs, and can react with complex outputs over time, often using memory to delay their effect. (By "genes" here, I mean the entire set of interacting brain structures, nerves, hormones, transmitters that are built through the activity of proteins described by the genes.) A "gene" by this definition includes "detection" circuitry, so it knows when environmental triggers occur.
Consider an imaginary "niche seeking" gene. Teenagers are always trying to find their own unique niche, to make them different from their friends. If two teens have the same "niche seeking" gene, they'll want to be unique from each other and avoid duplicating the other’s behavior. When one lifts his arm, the other (after observing this action) lifts his leg instead. When one goes outside, the other stays inside. When one rolls his eyes, the other one sneezes, just to be different. Same gene, different outcome.
Here's another example. The MAOA gene is known to come in multiple variants. Individuals having the AH variant of the gene are optimistic and resilient to bad treatment (in other words, they bounce back quickly after being treated badly). Individuals with the AL variant, on the other hand, react strongly to maltreatment, and often develop antisocial behavior in response to it.
So imagine two identical twins, both with the AL variant of the MAOA gene. Imagine that these twins were separated at birth, and one was adopted by a loving family (FL), and the other was adopted by an abusive family (FA). Clearly, the one raised in the abusive family would have a greater chance of becoming antisocial.
Although the ability to recognize abusive scenarios is shared in the twins (since how could a gene have any effect unless it recognizes the various scenarios for which it is designed to be affected?), only one individual received this environment. Yet both twins have the same innate branching circuitry. If situation A, then outcome B. If situation C, then outcome D.
Just because the outcomes are different does not mean that the behavior isn't 100% genetic. The environmental scenario detection is built into the design of the genes (i.e. the brain circuitry designed by the genes).
(Note, the full name of what I called the "AL variant" is "MAOA promoter polymorphism, low-activity". I just think all those scientific terms get in the way of understanding the philosophical implications!)