I have used Lehigh University Benchmark (LUBM) to test my application.
What I know about LUBM is that its ontology contains 43 classes.
But when I query over the classes I got 14 classes!
Also, when I used Sesame workbench and check the "Types in Repository " section I got 14th classes which are:
AssistantProfessor
AssociateProfessor
Course
Department
Fullprofessor
GraduateCourse
GraduateStudent
Lecturer
Publication
ResearchAssistant
ResearchGroup
TeachingAssistant
UndergraduateStudent
University
Could any one explain to me the differences between them?
Edit: Problem partially solved but now How can I retrieve RDF instances from the upper level of Ontology (e.g. Employee, book, Article, Chair, college, Director, PostDoc, JournalArticle ..etc) or let's say all 43 classes because I can just retrieve instances for the lower classes (14th classes) and the following picture for retrieving the instances from ub:Department

You didn't mention what data you're using, so we can't be sure that you're actually using the correct data, or even know what version of it you're using. The OWL ontology can be downloaded from the Lehigh University Benchmark (LUBM), where the OWL version of the ontology is univ-bench.owl.
Based on that data, you can use a query like this to find out how many OWL classes there are::
I'm not familiar with the Sesame workbench, so I'm not sure how it's counting types, but it's easy to see that different ways of counting types can lead to different results. For instance, if we only count the types of which there are instances, we only get six classes (and they're the OWL meta-classes, so this isn't particularly useful):
Now, that's what happens if you're just querying on the ontology itself. The ontology only provides the definitions of the vocabulary that you might use to describe some actual situation. But where can you get descriptions of actual (or fictitious) situations? Note that at SWAT Projects - the Lehigh University Benchmark (LUBM) there's a link below the Ontology download:
If you follow the "what do the data look like" link, you'll get another link to an actual sample file,
That actually has some data in it. You can run a query like the following at sparql.org's query processor and get some useful results:
I think that to get the kind of results you're looking for, you need to download this data, or download a version of the UBA test data generators and generate some of your own data.