I'll start by saying that my experience is in industry, not in a classroom. I worked for a fortune 500 company that brought a new division up to the billion dollar per year mark using an MS Access database. For five years, over a decade ago, I was one of the people working in Access daily to expand and manage the product until it was retired.
While MS Access has flaws and weaknesses, it's a real powerhouse in the right hands. Tuning an MS Access database is much like tuning DB2 without the tools. We can actually blame Access for the prevalence of SQL based solution in enterprise computing - IT departments using ISAM/VSAM/other "no sql" solutions couldn't compete with a business user at their desktop for cost/benefit for projects.
If you want to get enterprisey, all of the major players (Oracle, IBM, Microsoft) have free and education licenses. Of the three, MS is my own preference.
PostgreSQL is the only open source RDBMS that I recommend, mainly because it's the only one positioned as enterprise ready. MySQL and SQLite are positioned as entry level databases for projects that don't require enterprise features.
Edit: Comments prompted this realization that what I look for in a database for enterprise work is much the same as what I would expect educators to look at, and a list of those attributes might be useful, in no particular order.
RDBMS's are mission critical components that develop almost cult followings. While I tend to favor some RDBMS's over others, I have no emotional investment in favor of any I have used over any other.
1. Ease of Support - if a system requires a lot of support, it is expensive for the enterprise and untenable for the education market. DB2 is incredibly hard to support, so it is one I rarely recommend.
2. Consistency - if the RDBMS behaves in a consistent manner under a broad range of conditions, many quirks can be tolerated and worked around.
3. ACID compliance - Does it support "all or nothing" transactions spanning multiple statements? Partial commits are hard to debug. ACID compliance means easier error handling and better focus on the problem at hand. MySQL and SQLite fail this test, but are still fine for many types of problems.
4. Locking behavior - every RDBMS has is own way of handling latching and locking, which is best under specific circumstances. In my experience, generational or non-locking architectures are easier to work with in general because locks are rare events.
Postgres, SQLServer, and Oracle are least prone to locking issues. DB2's locking behavior is a consistent source of complaint from develoeprs and support teams. MySQL locks the entire database on every write.
Real life scenario - kids volunteer as developers with open community game servers and get good experience. I don't have the patience to tell the kid over and over that he's fixated on the wrong RDBMS for that particular job because MySQL locks the entire database for every write operation, and saving game state in a MMORPG is too write intensive for that product.
5. Standards compliance - none of the RDBMS' are 100% standards compliant, but the big three and postgres are gradually narrowing in on standards compliance. They mostly extend the standard versus violate it.
6. XA support - this allows products spanning multiple databases and even non-RDBMS data sources/sinks to use multi-phase commit logic to ensure the ACID compliance spans the entire system. Whether or not this is important depends on the system you are building. The big three and Postgres all have good XA support.
6. Support system - When you have a problem, can you find help?
In my experience, DB2 communities tend to be unhelpful ("why does't you company hire a real DBA?" is a common response to questions. IBM has education and free license options, but IMHO it's a pain to manage.
Oracle communities are very helpful - Tom Kyte is/was the product architect at Oracle, and his blog "Ask Tom" is one of the best sources of help for Oracle issues. Oracle's full standard database is freely licensed for non-production use. Oracle is difficult to set up initially, but very stable after that.
SQLServer is so pervasive now and is so easy to handle compared to the other "big three" that answers are usually a "google" away, usually on Microsoft's technet site or one of the Stack Exchange communities. SQLServer is easy to set up and requires almost no management once it is configured.
The open source databases tend to be very well supported by their developers and user communities.
Overall
There is no reason not to use the "big three" given the free and educational licensing options available. Of the three, SQLServer imposes the least disruption to the development (or education) process.
Postgres is right up with the big three for reliability and features. The first upgrade path for successful projects built on MySQL is usually to convert to Postgres. Since MySQL and Postgres are the same price, why not start with Postgres if you prefer to use an open source database? Postgres is (in my experience) less quirky, allowing you to focus on the matter at hand.