As I said in my last post (Over two years ago, wow), I feel that the disadvantage that Attack types suffer when launched first is disproportionately greater than the disadvantage that Defense types have when launched first. When battling against another blader, a good Attack launch will usually hit the opponent either almost immediately, or after a relatively short time. When launching both Beyblades yourself, the few seconds of delay between launches (especially with opposite-spin matchups, where a cap switch or launcher switch is necessary) is often enough for the attack pattern of the Attack type to atrophy, as well as Stamina and KOing power. Conversely, the Defense type, when launched first, suffers from a loss of Stamina (which is not usually an issue, though I would advise alternating launches if OS is a viable win condition for the attacker, like in the case of MF) and an increased chance of taking a direct hit, since it's a sitting target instead of a simultaneous launch. The Stamina concern is rarely a problem for the defender, which will usually OS by a significant margin, and the second problem is mitigated by using an indirect launch like the one th!nk mentioned (launching into a flower rather than directly against your target), which is more likely to emulate a tournament scenario where you miss the initial attack, but loop around for a hit.
It's also important to remember that testing results do not, and probably should not reflect tournament results. Blader skill, strategic choices, mistakes, and inherent randomness affect tournaments significantly; for test results, we want an idealized baseline that reflects the properties and performance behind the game, not the strategy and skill-based play of tournaments.
The existing system strikes a balance between accessibility and accuracy; its results do not perfectly reflect tournament results, but almost anyone can crank out 20 rounds with the appropriate equipment and have contributed to the community data set. If we were to emulate tournament performance, we would either need a testing group of equally skilled or randomly skilled bladers to do all tests, and would probably require a greater number of rounds to ensure accuracy despite all the mis-launches and other "failed" rounds. A bad launch proves nothing about a Beyblade's power.