The reason is that's what everybody else is up to these days. Curators are trying to hit the sweet spot where they front run large enough a vote mass to maximize their share of the curation rewards without forfeiting too much because of the early upvote penalty. Back in the day under #oldsteem it was easy to maximize your curation rewards by front running bid bots. That's because a) the typical bid bot voted late and b) the bot votes on the most highly rewarded posts were so massive compared to the rest of the vote mass that front running it for good curation rewards was easy.
That is no longer the case. The same methods that used to work under #oldsteem do not work under #newsteem. @abh12345 demonstrated in his excellent post titled HF21 - Bringing the earnings of maximizers and other curators closer together how the gap between the curation rewards per 1000 SP per week of the most efficient maximizers and that of the rest has got substantially smaller. When it comes to curation reward maximization by upvoting the most highly rewarded content, too many curators are currently trying to take the easy way to maximize their rewards. The fact is they end up trying to front run each other. I've even seen some former bid bot operating whales trying to optimize their vote timing by auto-voting some popular content creators at three minutes. Anyone smaller trying to vote after them will not see a great curation reward as a percentage of the value of their vote.
So, particularly if you're a Dolphin-sized user you should think twice about your strategy. What I recommend is you try and look for other authors to vote on. There are authors out there who already earn reasonably well but tend not to be as well-known as the best earning ones. Those people typically have a smaller number of powerful supporters whose votes you can try to front run. If you don't believe me, use SteemWorld. Open the Coming Rewards tab, activate Curation Rewards and hit refresh. On the left, you see pie chart icons. If you click on such an icon, a pop up window appears with the chart and a data table. You can sort the data by clicking on the titles. That gives you detailed information on who front run whom. Remember that the early upvote penalty runs linearly from 100% to 0% from zero to five minutes.
This is what you'd see. In the case of this post, the two largest voters voted over an hour later than the post was published. I voted at 4 mins 12 seconds and managed to front run all the largest votes. But my rewards were still nothing to write home about. My curation rewards were only 51.16% of the value of my vote. That's totally average. Despite the successful front running, there were simply too many about equally successful front runners I had to share the rewards with. It's clearly not a good maximization strategy to upvote highly rewarded material only. I'm not unhappy about voting on this post, however, because I like what OCD and the people who run it are doing.
But what I'm saying is that I love the way the curation reward system is starting to work as it was always supposed to. Resting on your laurels and lazily upvoting those who regularly get the highest rewards is not the way to get the highest curation rewards. It makes a lot of sense to go after the second tier. I mean those authors who are getting some predictable support you can front run but in whose case the entire usable part of the five minute window is not yet saturated with voters trying to front run each other. And when that tier becomes saturated with votes from opportunistic voters, they're going to have to find new authors to maximize their curation rewards on etc. etc. That's how support can spread from the top tiers to the lower tiers without "the good guys" having to fight a war against "the evil maximizers" for every inch.