Are course evaluation forms REALLY confidential?

Whether evaluations are anonymous (see StrongBad's answer for the difference with confidential, which evaluations are rarely) depends on your institution. On my university, they are. However, especially in small classes you can often get a pretty good idea of who wrote something depending on writing style and typical spelling mistakes.

But as Elizabeth Henning says in the comments, it's unlikely that this is the case here. It takes time before the evaluation is sent to the teacher. On my university, this happens only after all grades have been sent to the administration, after which it is possible to change them, but all those changes are logged - it may even be disallowed to mark someone down in that phase. In any case, the instructor won't care enough about the grade to change it and risk all kinds of trouble.

There is a difference between anonymous and confidential. I have never heard that course evaluations are confidential and I have seen faculty share evaluations. As for being anonymous, I have never seen an evaluation with a space for a name or student number, and the online systems I am familiar with do not make that information available. That said, the feedback, and sometimes writing, often makes it seem like you know who wrote the evaluation.

While you may be able to identify students, many departments try and keep the evaluations sealed until after the grades. Even in departments in which you collect the forms directly from the students, everyone I have talked to recommended not looking at the evaluations until after the grades were submitted.

While there is no benefit for marking a student down for their evaulation, if a faculty member wished to do this, in some cases they could. It would not be unreasonable to go talk to the director of teaching about your perceptions.

It really depends on your institution. Mine handled it pretty well: all evaluations had to be submitted electronically in a centralized system before the final exams to avoid retaliation from students, and were made available to the teachers after the final course grades were submitted to the registrar, to avoid retaliation by the teachers. We received printed results, where all numerical/rating answers were shown as total counts, and verbal answers were anonymized and put in random order. Except for answers citing specific personal interactions, it was impossible for me to make any inference about my students' responses. So at least where I worked, the evaluations were indeed as anonymous as it gets.