WorkBuddy summarizing a semester of club docs: one character off in a field name blanks the whole column—am I misunderstanding something?

I came across an article about large model selection, and one sentence stuck with me for a while. The gist was that 90% of teams choose the wrong model, mostly because they haven't tested it, and it has little to do with whether they understand the technology; the first principle of selection is that it's sufficient for your task and has the lowest unit cost, while the strongest actually ranks behind. After reading it, I realized I'd been using WorkBuddy wrong, but I hadn't thought about switching tools.

I've used WorkBuddy for two months, daily just throwing documents and spreadsheets in to organize, and it's been fairly handy. This week I wanted to summarize a semester's worth of club activity records into one summary, with about thirty materials—notification screenshots, sign-in sheets, Word files all mixed together. I created a folder and threw them all in, clicked the button on the interface that looked like automatic categorization, waited a few minutes, and the output wasn't useless, but basically couldn't be used directly.

The weirdest thing was the date column. Some entries were empty, and some had dates written wrong. I thought the screenshots were blurry, re-uploaded them, and it was still the same. Later, I processed those few separately, and surprisingly it was correct. So I guess the problem is when processing together, it mismatched fields from different sources—nothing to do with recognition.

There's something even weirder. At first, my header was "Date", and in later sheets it was "Time". When merged, that part the entire column was empty, not a single entry. After changing both to "Date" and running again, it was normal. I only realized this after seeing someone else's post—turns out field names must be exactly consistent; it won't automatically align them for you, and if you write it wrong, it just shows empty.

So now I'm stuck here. How exactly should this kind of multi-source material be fed—throw it all in at once and let it categorize itself, or should I unify the format first? I tried unifying first, and it was indeed much more accurate, but manually unifying thirty-some files was slower than my original manual summary. Is there some setting I didn't enable, or does this kind of work inherently require cleaning first? Friends who've actually run this successfully, can you share how you did it?