Based on what you described, I think splitting positive emotions (like enjoyment) and negative emotions (like anxiety and boredom) into separate analyses makes more sense. Combining them can cancel out opposing effects, leading to a non-significant result, even if individual studies show clear findings. Positive and negative emotions are distinct, so analyzing them separately will give you a clearer picture of the intervention’s impact. Plus, it’s a common practice in meta-analyses.

In the data you are talking about, these parameters are highly dependent on each other, so I think it would be more correct to look at them separately.

Positive and negative emotions have different predictors and also predict other things.

This text tested and demonstrated what you are finding in different experiments: Article The independence of positive and negative affect

This other text discuss the case of valence: Colombetti G (2005) Appraising valence. J Conscious Stud 12:103–126

We recently published a research comparing emotions and emotions moderates differently in different countries (we compared France and Brazil): https://linkinghub.elsevier.com/retrieve/pii/S1162908824000938

I am just describing this because emotions could have a cultural effect as well. I don't know if you control this possible effect.

I hope it helps you.