Now it's starting to make sense, but the book's approach seems like more work than is necessary. If you already know that a horizontal asymptote is y = 1, creating a new function y = f(x) - 1 will have the x-axis (the line y = 0) as a horizontal asymptote, as you said.
\lim_{x \to \infty}f(x) - 1 = \lim_{x \to \infty}\frac{x^2-1}{x^2 -4} - 1
=\lim_{x \to \infty}\frac{x^2-1}{x^2 -4} - \frac{x^2 -4}{x^2 -4}
=\lim_{x \to \infty}\frac{3}{x^2 -4}
Now I understand where the 3 came from.
Since the numerator is positive, and the denominator is positive for all x > 2, the value of this limit is 0, and the limit value is approached from above.
This means that the limit of the untranslated function is 1, and that the graph approaches the horizontal asymptote from above, as well.
To find the behavior as x approaches negative infinity, evaluate the same limit, but with x approaching negative infinity.
As already stated, this seems like more work than is necessary. By bringing out a factor of x2 from the numerator and denominator, you get the horizontal asymptote directly and you can tell whether the graph approaches it from above or below.
\lim_{x \to \infty}\frac{x^2-1}{x^2 -4}= \lim_{x \to \infty}\frac{x^2(1 - 1/x^2)}{x^2(1 - 4/x^2)} = \lim_{x \to \infty}\frac{1 - 1/x^2}{1 - 4/x^2}= 1
For a given value of x, we are subtracting 1/x2 from 1 in the numerator, and are subtracting 4/x2 from 1 in the denominator. IOW, we are subtracting a larger amount from the denominator. This makes the denominator slightly smaller than the numerator, which makes the overall fraction slightly larger than 1. This is the same result we got earlier, but is more direct.