You lose many of the benefits of using a repository if you stick to just one branch. It can become tricky to identify which files were modified by which features, and while you can rewind any file at any time, it starts to get dicey when you're rolling back changes to specific files to try and find a specific point when a bug was introduced or to roll back a feature. You're more likely to screw up your repo with this strategy than you are to "forget" to deploy everything. I wouldn't recommend or wish this strategy on my worst enemy.
Are there any disadvantages of that beyond being a bad practice and having a mess in the repository?
You say this like it's a trivial matter. It's not. The odds of things going wrong in your repo are potentially far more catastrophic than "forgetting" a file that you can then deploy later.
The strategy of using a single branch is like driving a car without car insurance. The more you use the car, the more likely it becomes that an accident will happen. If an accident never happens, you "wasted" a bunch of money on insurance, but if you happen to need to file a claim, then the insurance will have more than paid off. Contrariwise, if you have an accident without insurance, it could be potentially catastrophic.
Similarly, you may downplay this strategy in the short term, but the longer you do this, the more likely you are to one day have an accident and be unable to recover easily, or possibly even at all. It's simply not worth the risk.