The unicode codecharts specifically state that the sample glyphs shown are not intended to be normative, but I'm wondering how much leeway is actually intended. I'm specifically interested in symbol/picture characters rather than characters from actual scripts.
As a concrete example, both Linotype Astrology Pi No. 1 and Linotype Astrology Pi No. 2 include glyphs in the range U+2648 through U+2653 (i.e. zodiacal symbols). The glyphs in LAP#1 closely match the symbolic representations shown in the standard whereas those in LAP#2 contain pictorial representations of these signs which bear no resemblance to the glyphs shown in the standard. This seems legitimate, though, since each set of glyphs match the descriptions given in the standard (aries, taurus, etc.)
However, I'm wondering about cases where the description given in the standard describes an actual _shape_.
For example, in the standard U+26E8 is described as BLACK CROSS ON SHIELD (=hospital). Would it be inappropriate to encode a glyph in this slot such as the hospital glyph from Adobe Carta (encoded in the private use area at U+E02D) despite the fact that it is actually a white cross on a black circle?
Note that the above is just an example. I'm just trying to get a feel for whether the graphic description takes priority or whether a graphic or semantic match is sufficient (or near-match for that matter - with pictorial symbols in some cases the semantics are a bit fuzzy).
n.b. I'm pretty sure Adobe Carta was released before this codepoint had been defined, so their decision to use a PUA codepoint is not revealing of their take on the matter.