Isis the Scientist had a post yesterday (with lots of discussion) about in what order the various sections of a lab report should be written. Whatever. People, do what you want, just make sure that the final product is good. To me, the most important part of writing a report occurs before you start experimenting: it's the the time when you make up the data.
Yes, make up the data. Believe it or not, this was something I learned to do as an undergraduate engineering student, and the professors actually encouraged it.
Hear me out. The junior level chemical engineering lab was a royal pain in the posterior. It was long, it involved larger pieces of equipment than any of us had ever worked with and we were given little guidance as to how to run it. Each week, the team would receive a memo written out as if it were something that your supervisor (in industry) might send your way - the overall objective was clear, but there was little guidance as to how to proceed. Days before the lab started, we had to look at the equipment, imagine how it would work and also figure out which equations we would use to analyze the data. We would then have to make a proposal to the prof as to what our plan was and what data we would collect.
Now while we all (thought we) understood the equations being used, the biggest challenge we faced was making sure that we collected all the data that we needed when we did finally got into the lab and used the equipment. If we didn't get the data at that time, we were up the proverbial tributary without a proper means of forward locomotion. We would not be given a chance to back into the lab and rerun anything (much like industrial situations, by the way). So the best way to avoid the problem was to make up some data before we got into the lab, and run some calculations with it to make sure that we had all that we needed.
This actually provided us with positive two outcomes: first, we could verify that we were going to be collecting all the data that we needed, and second, by making what we thought were realistic guesses for the variables (as opposed to random numbers), we were training ourselves to anticipate appropriate values. If the actual values were significantly different, then we were able to learn why out hunches were so far off.
It's still an approach that I use (mentally) to this day - think first about the all the data that you will need and how you will get it. You may not get the results you want, but you will be sure that you can stand by what you do have.
Just make sure that you don't confuse your made up numbers with the real data.