The actual command that runs if the if condition is true makes no difference, as long as they are all exactly the same for each test
This is exactly what I tend to consider questionable: even if they are exactly the same commands, fluctuations in execution time can be introduced, and they are greater as those commands get more complex (for example evaluate many variables/identifiers etc). In a benchmark in assembly or even C/C++, this would probably not matter, as a single command takes a predefined number of cpu cycles. In higher levels though, like mirc scripting, the same command can be completed in times that vary among repetitions.
This doesn't matter much if you want to measure the speed difference of two procedures that one of them is known to be faster. But I think it matters when these procedures are almost equally fast. In that case, the 1 msec (for example) difference between those procedures might be lost in fluctuations during the evaluation of variables or what not. Granted, a very large number of repetitions finally evens out these fluctuations, but it feels more proper to try to measure the time consumed by only a specific procedure (to the extent that it's possible) than including the execution time of other elements and then ignoring them as "common factors", which might not be so "common" after all.
