The emails are randomly split between templates in a given step. However, outcomes are measured and displayed per template (as long as you link a template, not clone it) to help determine what language is most successful in a step.
Moreover, A/B testing, or creating alternative templates, is not the way to communicate to differing targets. So, keep your message consistent between templates in the same sequence.
To test a separate messages or communicate to separate targets, create a new campaign! You can even clone and adjust a sequence for a new purpose.
A/B test your templates in each sequence to improve your communication to each target market separately.