Yeah - it's a very complicated issue with multiple points of view embedded in the texts themselves and then we are translating them with different purposes that aren't always explicit.
In the past, I personally had attempted to be very literal and keep to what the Chinese text says, as far as it could be discerned, without attempting to read an Indic original through the Chinese. I still remember once getting sniped by an academic for translating 世尊 as "World-Honored One" instead of something like "Blessed One." Apparently it was annoying that I had not reverse-translated the Chinese to Bhagavat first and then translated the Sanskrit to English. I countered that apparently the Chinese translator thought it meant 世尊. Then he pointed out I really ought to go to school and get a degree. So, if I get a PhD (which doesn't make a person an expert in much of anything), I supposedly will be able to edit the Chinese translator's text for him 1700 years later. I don't have the original he was working from, so how am I supposed to do that accurately? But translation methods in the Buddhology world are a little different than the rest of world. I mean, we've academics creating whole new hybrid texts from two or more Chinese translations performed centuries apart in time. How that makes them versions of the same text is beyond me. Yes, it's interesting, I guess, to represent several texts and their evolution in parallel on the same page. But is that translation, or authoring, of Buddhist texts?
Now, myself, I have become very interested in early Chinese translations as a record of the Central Asian Buddhism that has since disappeared with scant remains to study. In that context, it does make sense to try to read through the Chinese where it is not terribly uncertain. But its such a blurry thing to attempt to reconstruct the original terms when they were rendered quite differently from one era to the next or from one translator to the next. It's very time consuming to try to build a lexicon for particular translators - which is what is required if we are to be as accurate as possible and not simply "make it up" as we go. And that is only possible indirectly with a few translators, like Kumarajiva or Xuanzang, using Sanskrit manuscripts that are not the same ones they were translating.
So, yeah - it's a big mess.