The other answers are good and cover everything, but I wanted to add one to try and be crystal clear in highlighting what you are missing.
By default there is a SPECIFIC harmony to use for each degree in roman numeral analysis (RNA), unless specified otherwise. The harmony, ie. which chords each degree of the scale specifies, can be major or minor. In a minor key the RNA will place the tonic minor on the i, a major key will place the tonic major chord on the I. Note that in minor keys the existence of 'natural' and 'harmonic' minor versions could indeed introduce ambiguity on the quality of the vii scale degree, but rest of the RNA system is NOT ambiguous.
Roman numeral analysis expresses chord relations without NEEDING to specify a key (although saying/being clear it is of type major or minor should be a minimum), but a key must be chosen in order to play it. It is useful because it describes the harmony of a piece in a way that can be applied to any given tonic root.
For example, a popular jazz tune has chords;
ii, V, I, IV, vii, iii, vi
Knowing this piece expressed in RNA I can play it in any key very easily and without having to transpose. In fact, knowing songs in this way rather than explicit chords is highly recommended, and in certain genres (jazz, some pop) a required skill. I appreciate that may not be your goal, but for context.
If I learned a song in roman numeral analysis then I know it in every key immediately, I just centre the I chord around the given tonic and play the relationship. This is incredibly useful for both analysis and performance. Analysis cares not what the starting note is, more the complex relationship between notes. It helps us, a little like algebra, to look at the generalised relationship and behaviour of a system rather than one specific example.
Also, each chord in roman numeral analysis tells me the 7th quality of the chord, and the further upper extensions that are available, it is much more than just knowing which chords are major and which minor.
Expressing the Andalusian cadence in RNA using the minor tonic as i gives us i, VII, VI, V. You could also express that chord sequence against a major key as vi - V - IV - III7. The second way may be useful if that particular chord progression appears in a piece that otherwise can be considered to be in a major key.
RNA is incredibly useful, it can demystify complex sequences of harmony in a way that would be tricky to spot from a chord sequence alone, it gives us information about which chord extensions will sound most consonant, it allows us to change key with ease, it allows us to see compositional tools used across various sections, keys and pieces for comparison and greater understanding.
If you are familiar with programming, using chord symbols is like 'hard coding', very explicit and clear, but rigid and not that translatable. Using roman numeral analysis is learning to understand the dynamic system we are operating in. It means that sometimes there are not hard and fast rules, but an expert on the system will quickly recognise the patterns and could express how it would most likely be interpreted.
The examples given in the wikipedia article are not that useful, there is rarely cause to respell RNA against a modal root. It is perhaps of historical curiosity if you are comfortable with modal harmony and have a strong basis in RNA, but I can see how it could be very confusing. Once one is comfortable with using RNA and can see why the Andalusian cadence is usually going to be i VII VI V etc. then the counter examples in the wiki article become very easy to comprehend, not in danger of destabilising the use of RNA, but also of very limited use. Though perhaps they could inspire new compositions, any alternative viewpoint can be the basis for a new exploration. For this reason it's important not to regard theory as an absolute ruleset, to be adhered to at all times. However this doesn't mean that the well established systems we have are not very thorough and complete tools.