Not wrong, but different.

I don't know what `em' in web browser means, however, in web browser,
for example, it is possible to create an image of 3em height and
display it like this:

.----------.
|          | Text line 1
|          | Text line 2
|          | Text line 3
`----------'

And image will be displayed solely and will fit into 3 lines height.

In Emacs, if image with 3em in height is displayed using slices, it
will result in a gap in the final slice, because 1em (font's size) is
less in pixels, than 1ch (font's height and final line height).

Let me elaborate: For example we have 45px font, with ascent 38px and
descent 10px, resulting in font height equal to 48px.  And we create
3em image, which will be 45*3 = 135px in height.  Now, we display this
image using slices: first slice will be: 0,48 (to occupy full line
height), second slice - 48,48, and third slice 96,39.  Last slice is
39px in height and if you display it with :ascent 'center it will have
4px gap.  If you display it with :ascent 100, it will increase final
line height by 1px, because 39-38.  You probably can calculate correct
:ascent value to get exact value of ascent in %, but, unfortunatety,
ascent is integer and in general integer number of percents is not
enough to get exact value (for example for fonts larger then 100px).
Also, this forces to modify :ascent value for the last slice, which
leads to additional garbage.

`ch' and `cw' specifiers solves two problems:
1) Makes it possible to create image intended for slices display
without gaps, and automatically adopts to text-scale changes.

2) Makes it possible to align images precisely to the width of the
given number of characters, especially this makes sense when monospace
font is used.  Also adopts to text-scale changes.