having a dataheet as a resource and then comparing values dynamic values

For your first question I've often wondered if we should have a tool to create a table like that. What you can do is put a Creator and AttributeManager down, and in the AttributeManager create an attribute whose value is a csv format table (use the text editor to do that). Then you could use the AttributeSplitter to create a separate feature per row, and again to split the columns into attributes. I think that should work, although I don't have FME at the moment (just installing a new version).

For your second question, you could use the FeatureMerger or FeatureJoiner transformers, or - if you prefer using SQL - there is the InlineQuerier transformer.

I'd lean towards keeping it in a CSV file external to your workbench, you can then read it in when you need it - I've done similar things like this, you need to make sure its documented that the workbench is dependent on the external CSV.

Alternately, and similar to things i've seen others do, is store the data in a PythonCreator and have that create the features you need. Something like the below in a PythonCreator will work. Note you will need to expose the fields you create

import fmeobjects
 
class FeatureCreator(object):
    def __init__(self):
        self.fields = ["f1","f2"]
        self.data= [["v1","v3"],["v2","v4"]] #store the data row by row
    def input(self,feature):
        for row in self.data: #iterate over each row
            newFeature = fmeobjects.FMEFeature() #create new feature per row
            for value,field in zip(row,self.fields):
                newFeature.setAttribute(field,value) #set field and value for each feature
            self.pyoutput(newFeature) #output feature
    def close(self):
        pass