score:11
Romans used both additive and subtractive and this is true from late Etruscan times up to middle ages (1100 Arabic numbers spread into Europe).
Roman numerals were never a stardandized notation system, but a rather simple tally-mark system that got complex on some use-cases. There were variations that depended on context and user (writer). For instance, the notation of large numbers over thousands was crazy varied.
Regarding subtractive/additive, the numerals for 4 and 9 only had two options each, but consider that the numerals for 17, 18, 19 when written had three options each (18= decem et octo, duodeviginti, octodecim). All three options are still used in different Romance languages. So, their use never did stop.
Here is a list of major factors on popularity for each system:
It is true that IIII is the simplest and earliest form to write 4. During early Roman times the additive form is preferred, but we don't know for sure wheter it was exclusive, since you will find IV in quite early republican inscriptions in the CIL.
The theory of "avoiding Jupiter wrath" by not using the abbreviation IV, in my opinion, makes no sense. The ancient and earliest Roman (read: Etruscan) name of Ivpiter was Iovis, so there would be no clash and no God's wrath by writing IV in early Roman times. It just makes no sense. The "simplest tally mark" hypothesis is more valid in my opinion, by Occam's razor argument.
On another topic, the reason that clocks use IIII and not IV might be to avoid confusion with VI when numbers are written aligned along the perimeter (you can write both inwards OR outwards along the perimeter), and not vertically aligned. In fact both clocks with inward and outward Roman numerals exist. Big Ben is inward, most watches are outward. Yet, this is only my personal opinion. I see that IX and XI face same problem.
To learn more about variety in literature use and regions, see: Subtractive Versus Additive Composite Numerals in Antiquity