I am not trying to be rude, but I would have to disagree. Most researchers head out with a goal in mind, and a general understanding of what will be involved. Consider the atom bomb, the researchers wanted to make an atom bomb and that is what they got. Consider a lightbulb, though Thomas Edison didn't know all the components, he knew most of them and in the end he got a lightbulb and not a fusion reactor. Consider cancer research today, I would be really surprised if a scientist exclaimed "I discovered teleportation technology" while researching cures for cancer. Consider the automobile, the airplane, the artificial diamond, the spaceshuttle, the computer, the television, ....
Yes, but there was a process leading to each of those; and early enough in those processes, no one would have thought "I'm going to try to make a lightbulb!" or "I'm gonna make an atomic bomb!" So really, it does boil down to what Ynglaur said:
Also, in terms of "real world" research, there tends to be a difference between scientific theory and engineering, and the development for each is often--though certainly not always--handled by different people.
Except not only are the two often handled by different people, but the engineering typically comes after the scientific theory has been hammered out. And there is quite a bit of randomness in the hammering out of scientific theories. For example, the idea that one should theoretically be able to produce a nuclear chain reaction first occurred to Leo Szilard on a street corner, contrary to Ernest Rutherford's conclusions and the general consensus of the time. He was basically daydreaming about physics when the idea struck him... The notion that one can create a nuclear chain reaction must exist before anyone can set out to create an atomic bomb.
Also, the invention of the computers that don't occupy a warehouse was a side-effect of the Apollo missions. They needed computers to handle all the data, but computers of the time were way too big to send to space. So they devised a way to build much smaller computers without sacrificing functionality. Cancer research (and medical/drug research in general) produces results that have wide-ranging impacts in broader fields. Obviously it won't lead to teleportation, but it does often lead to advances in other related fields.
If you tell your research advisors that you want more durable swords, there are many paths you might go down. Maybe your researchers will devise a method of forging the same materials into a stronger alloy, or maybe they'll figure out how refine some other type of ore that can be forged into a stronger metal.
If you give your researchers a goal to aim for, then the variety of what you might end up with should satisfy that goal. If you ask for improved swords, you shouldn't discover a better way of managing your fields. If you demand only that your researchers work on methods of improving infrastructure, you might learn how to build better roads, or walls, or wells, or whatever - and maybe an occasional something that doesn't fall under 'infrastructure improvement' but is related in some way. Going back to the weapon example. Assume at the beginning of the game you task your researchers with the goal of improving your weapons, before you've figured out how to make metal tools. In the process of figuring out how to make better weapons, maybe they'll figure out how to forge metal tools; but even before they perfect it to the point of making useable swords and axes, they might get good enough at it to produce metal farming tools. This is completely in line with the historical process of invention and I think something like it would make the game quite interesting.