I do agree in-person may be best for initial relationship-building, if possible!

But...

Mehrabian's study was limited in methodology and applied only to communicating emotional states, not communication of information in general. He spent a lot of time campaigning against the way his results were popularized and repeated as %-based rules about communication in general.

The findings are not necessarily applicable unless:

The communicating parties have misaligned incentives or some other reason to be deceptive about their own emotional states.

Those hidden emotional states are extreme enough that hiding them will impact the potential for productive outcomes.

I don't think most workplace conversations will resemble hostage negotiation in these ways (at least, I hope not!). If we're just talking about clarifying questions, conversations with low emotional stakes, or communication where there are shared incentives or a baseline of trust those criteria aren't likely to apply.

In some cases even the "lost" information may be more noise than signal (on video it can seem like people aren't giving eye-contact, for example) and most people can't actually reliably interpret body language signals with significant accuracy. And I can also imagine cases where slight deception about emotional states might even be a benefit for productivity (Ex: disguising the fact that you are bored or unimpressed favor of acting polite or encouraging)