First two things to consider is inplace animation is reactive while root motion is event driven. For inplace to work it needs something to react to such as the movement driven by the movement component so what is actually moving is the capsule and the resulting pose is based on the current state of the capsule. In order a state is changed when activated, by key being pressed, and the animation is played based on the current state.
Root motion is activated as part of the event of a key being pressed and since the animation is built into the animation data does not need the movement component as a required element to react to.
Next
When you tell the animation BP to “use root motion from everything” all input as to driving the movement component no longer function, as root motion is “event driven” how ever the capsule is glued to the root, for basic collision, as well makes use of the basic physics of the character always falling. If you make use of RM then most if not all tutorials based on inplace will fail to work as expected as well network considerations needs to be considered as the requirements are driven in part via the movement component.
The basic RM movement system.
