Addendum

I noticed that some papers in my library from Springer publications also had this issue and the Pages field was populated with a page range (e.g., “1–n”) instead of the article number. However, I tried saving one of those papers again, and now the Pages field is correctly populated with the article number. For example, I tested this paper:
https://link.springer.com/article/10.1140/epjb/s10051-024-00784-3

Despite the webpages for Nature and Springer journals being virtually identical, they use different translators:
• Nature: https://github.com/zotero/translators/blob/master/Nature Publishing Group.js
• Springer: https://github.com/zotero/translators/blob/master/Springer Link.js

In particular, the Nature translator includes the following code, which is not present in the Springer one. I believe this is responsible for the behavior I’m seeing:

if (item.pages && !item.pages.includes('-')) {
var start = text(doc, 'span[itemprop="pageStart"]');
var end = text(doc, 'span[itemprop="pageEnd"]');
if (start && end) {
item.pages = start + '-' + end;
}
}

if (!item.pages) {
// For some online-only publications this should be the "Article number"
// but is not present in either EM or RIS. We grab it from page
var page = doc.querySelector('.citation .page');
if (page) {
item.pages = ZU.trimInternal(page.textContent);
}
}

My understanding is that the “Article number” is now included in the RIS file, at least for the journals I’ve checked. Because this value doesn’t contain a hyphen, the first if block is triggered, resulting in an artificial page range being generated.

@AbeJellinek
Thanks a lot for fixing this. My apologies for not respecting the etiquette of bug reporting.
I have indeed tried with a couple npj journals and it works as intended now.