As long as your errors are randomly distributed you should keep making observations until you get an error of -347% in the other direction, thus cancelling out your previous error and allowing you to model your entire existence with Riemann curves

No measurement is perfect. If your statistics are well executed, the measurement is a valid contribution to human knowledge.

Also every measurement is of great impact as long as nobody else can measure more precisely. So just wait 😉

Back in my early years of uni, we had one of those tests in which every question depends on the previous answer and we had to round every answer to two significant figures. For some stupid reason, a friend and I decided to also use the rounded figure when inputting the values into the next question instead of working algebraically. By the last question, we were off by like 5.

One time my lab partner kept the waste and threw out the product

“Clearly the model is wrong”

The main difference between Science and fucking around is that Science has you write things down. So long as you record everything you did, even if it ended up with wildly unexpected results, it's valid.