Unfortunately Probe blocks within library models do not work as they do in a non-library subsystem, and will lose their associations when the library model is included in a new model. This may be addressed in a future PLECS release, but the intended method for Probe signals associated with library components is to use Mask Probe Signals. The library model file needs to have a custom Subsystem (it can have additional blocks too), which can then be masked, and in the Subsystem Mask Parameters window there is a Probe Signals tab where you can define any internal signals you want accessible by a probe. In this way, you can probe the actual subsystem/library component itself when you drag it into a new model, and then monitor any of the signals you set up.
Let me know if this is clear.