To be fair, there are a lot of sweeping theories and conclusions in history, archaeology and paleontology that rest on some very thin reeds. The further back in time you go, the more it's like trying to figure out what a 1000 piece puzzle looks like based on a half-dozen random pieces. The more pieces you find, the better chance you get that something will fit together and have a decent chance of being a good sample of the whole, but making definitive, generalized conclusions from such paltry, fragmentary evidence is dubious at best, and very suspect to influence from whatever preexisting biases you brought with you. It's the blind men and the elephant.
Science and engineering is easy. You want to study something and observe it's properties? Build it, run an experiment, and do it. Do it a dozen times, or a hundred. You have the luxury of amassing a supply of data that is orders of magnitude better than what an archaeologist could collect from a lifetime of digging up middens and graves, so there's far fewer blanks to fill in with supposition and conjecture.