Why do people say this? This is a sport with a rule set that is agreed upon by BOTH fighters coming into the fight. It doesn't matter who you think won the fight as a whole, what matters is what you feel the scores are round by round.
Well if the scoring system is applied correctly, 10-9s don't have significant disparity from one another (i.e. one 10-9 is about equally dominant for the winning fighter as another) and ditto for 10-8s, 10-7s etc.
When that's the case, the round by round and the overall score roughly even out.
I've suggested before that the best scoring system is where the first round is scored such that the remaining rounds can be scored relative to it. So the first round, if it's 10-8 let's say, sets the definition of a 10-8 for the rest of the fight. A round about halfway as dominant as the first is a 10-9. A round more dominant by the same halfway margin is a 10-7. A round twice as dominant still is a 10-6.
If the first round is won by someone by a margin slim enough that there's no significant intermediate step, it's a 10-9 and again it defines what one point is worth. So the next rounds are compared to it, and scored accordingly.
An even first round (or, I've also argued, a round that's predominantly a "feeling out" round where nothing significant happens and neither guy is taking clear steps to win it) would be a 10-10. If the first round's even, it's scored 10-10 and the second round plays the role the first normally plays.
A first round that's so dominant that you can see two clear intermediate steps between it and an even round is 10-7.
Generally, a first round with a clear winner where the judge can compare the round to an even round and see a need for an intermediate score would be a 10-8, and this would be one of the most frequent scores for the first round. Which is why I used it in the first example.
I think this fixes two big problems:
First, the fights where two rounds could go either way and the other one is clearly won by a fighter (but not enough to make it a 10-8 under current scoring) are some of the most common bad decisions, because two judges have the same guy winning both close rounds, and even though the other fighter clearly won the overall fight, he loses. In such a fight with my scoring, you'd have at worst a draw.
Second, and most importantly, it makes the fight scored by round have almost the exact same result as scoring the fight as a whole, but would eliminate the "recency bias" (over-weighting action that happens later in the round) that used to result in bad judging when fights were judged as a whole. This is most important because that's what a good numerical scoring system does -- it makes the round by round score more analog, making the final score more like a well-judged fight scored in its entirety.
The problem with the current scoring is that it's too digital, and effectively binary with 10-8s being so rare (and 10-10s rarer still.)
Which reminds me of a third advantage: the fear that 10-8s or 10-10s will result in draws is mitigated by the fact that we use a more diverse scoring range, so we aren't as married to odd-numbered scores to have decisions.