Likeability gain from Conversation

Page currently says:

"Speaking to a squire about the topic they want to talk about raises their likability experience by 1% if their likability level is 1-2 or 3% if level 3-5, talking about a neutral topic has no effect other than using up one daily interaction, while speaking to them about a topic they currently dislike will slightly reduce their likability experience by 0.3%."

I'm currently training Elsie. I can confirm the gain was 1% at level 2 and 3% at level 3. However, at level 4 it is only 1.5%, not 3% as claimed above.

I have trained Logan, Kanna and Dai to Level 5, but that was so long ago that I can't recall how they behaved at level 4.

So, does anyone know if the paragraph quoted above is wrong for all squires at Level 4, or is Elsie behaving differently to the other squires at Level 4?

The % can change between levels and squires because the total points needed for that level can be different. Selecting the right option would be +3 points. If the total points needed for that level was 100, this would be 3%. If the total points needed for the level was 200, it would be 1.5%. You can see the flat amount of points needed by hovering over the bar. The paragraph speaking purely in percentages is kind of misleading, as the various squires have varying amounts of likeability per level, though they all seem to be on a scale of out of 100 at level 5. But a wrong answer is always -1 point and a right answer is always +3. Whatever that translates to in a percentage is based on the squire and their current likeability level, but hovering over the bar will still show how many points you actually have and still need.

Thanks. Understood.

So to make this page clearer, what we really need is a two way table showing the total number of likeability points at each level for each squire. For example, for Eirlys, Level 1 has 70 points.

I've added the table, with all the data I'm sure of, so it's still got a lot of unknown values. I've also rewritten the squires section, using ideas from lyasenu's reply above, so hopefully future readers won't develop the same misconceptions that I had. And I see Whitepass has already fixed the dodgy section quoted in my first post in this thread.