Asking whether something is or is not a conflict of interest is generally a misleading question, since it suggests a sharp cut-off rather than a continuum. It's very rare to be so completely detached as to have no problematic incentives (for example, many other authors are friends or rivals). Instead, the question is how troublesome the incentives are and how to handle them.
Recommending that the author cite your own papers certainly has the potential for abuse. However, people generally don't worry about it very much, since the worst-case scenario of a few borderline unnecessary citations is not so terrible (if the referee asks for many citations or they are really not relevant, then the editor should notice and intervene).
What's tricky about it is that some papers must be cited to give appropriate credit. If you discovered X, but the author mistakenly credits it to a later paper based on your work, then you have an obligation to tell the author about your earlier paper. The difficulty is in distinguishing what must be cited from what you wish would be cited. I don't think there is any standard rule for how to make this distinction.
What I do in practice is to make explicit recommendations to cite my work only when I can give a clear factual reason. I.e., if the author says something historically incorrect or gives the wrong attribution, then I correct it. However, I'm reluctant to ask the author to include a citation if everything in the paper would be correct without it. For example, sentences like "Previous applications of Smith theory include X, Y, and Z" are still correct even if I wish W were also included.
On the other hand, the next best thing is to say something like "The author may find the following related papers interesting: ..." There's an implicit suggestion that citations could be appropriate, but it's less pushy than saying it explicitly, and the author can decide whether to include citations.
This doesn't really solve the conflict of interest issue. After all, the implicit suggestion is pretty transparent, and the author may still feel some pressure. However, it seems to me like a reasonable compromise.