There's a reason carnival barkers hawk the guess the weight game. Height and weight are extremely difficult to estimate. With weight, a lot of physical factors come into play...muscle weighs more than fat, bone structure often gives a false impression of weight, a woman with very large breasts and a small waist may weigh much more than she appears to, and so on and so forth. With height, perspective and reference points come into play. Was the person far away, towering over the witness, did he/she have poor posture, was he/she observed stationary or moving, etc. etc. Add to both of those things a stressful situation that is rapidly unfolding and it's not difficult to understand why witness descriptions are often highly inaccurate.
There's a show on NatGeo called Brain Games. On one episode, they created a scenario of a car chase that resulted in a crash. They staged this in front of eyewitnesses, and then asked the witnesses questions like what color was the car, how fast was it going when it hit the second car, how many people were in the car, and so on. How the interviewer worded the questions often lead witnesses to recall events differently, even though they had all seen the same thing. If the interviewer asked, "How fast was the car traveling when it smashed into the parked car?", witnesses increased the speed. If the interviewer asked, "How fast was the car travelling when it bumped the parked car?", witnesses estimated the speed of the moving car much slower.
Many things come into play with eyewitness accounts. It's wise to always take these accounts with a grain of salt, imo.