Storyline Text-to-Speech loses neural voice when using hyphens to separate letters

I work for an organisation that has an initialisation in its name and we often create courses with text-to-speech. We use a neural voice (English UK: Amy). The initialisation is also a natural word so I have to specify to TTS to spell it out.

A while back I noticed that separating the letters with spaces created too great a pause between each letter, so I started separating them with hyphens, which sounded more natural (e.g. 'D-O-G' rather than 'D O G').

More recently however, any time I try this, the neural voice is replaced by the standard voice.

I understand that using SSML tags (<say-as interpret-as="characters">) will cause the neural voice to drop too, so I wonder if hyphens are interpreted as a shortcut to this somehow? But I still have older projects where I use hyphens and this doesn't happen, so I wonder if it's a Storyline issue.

Example:

Neural voice dropping | Review 360

Example of it working fine in an older project (press the speaker symbol to activate):

Neural voice not dropping example

Story file attached.

Storyline version is 3.99.34516.0