EDIT: With some more research, I found the reason. As I expected, it's due to historical unification of JIS X kanji with Unicode codepoints. Here's a GitHub thread about the issue, including precisely the two characters you asked about.
The tl;dr is that the Japanese codepoints corresponding to those characters, which were unified with the Simplified Chinese characters, were narrow to begin with, and so have stayed narrow in Japanese rendering in order to stay compatible.
Check out Ken Lunde's answer:
The Japanese (JP) forms of U+5173 (uni5173-JP; CID+11297) and U+590D
(uni590D-JP; CID+14484) are intentionally narrow (they correspond to
Adobe-Japan1-6 kanji, which in turn correspond to JIS X 0213:2004
kanji, specifically 2-03-08 and 2-05-27, respectively), and are
referenced only by the Japanese and Korean subset OTFs and Japanese
and Korean OTC font instances. (Note that these characters do not
correspond to hanja in KS X 1001 nor KS X 1002, so the Korean subset
OTFs and Korean OTC font instances are merely inheriting the default
glyphs for Japanese, which is by design and intentional.) They are
also the default glyphs for the multilingual OTFs because Japanese is
the default language. If you check other Japanese fonts, you'll see
similarly narrow glyphs for these two characters.
As to why: I could only venture a guess -- perhaps they were construed as character parts in Japanese digital fonts; since these characters do not properly exist in Japanese, just like 艹 or ⺝, they therefore don't have a "proper" square rendering.
The CJK Radicals Supplement section of Unicode has many other radicals whose rendering doesn't quite look "square", because they are used to describe parts of full characters, without being full characters themselves.