Recognizing spam from a "ham" (the term used for non‑spam) is an inexact science, and OVH is no exception to the rule.
So there are false positives.
However, OVH has a very inelegant way of handling spam in the outbound‑mail process. If it is detected as spam, the mail in question is simply deleted. No one is notified.
Nevertheless, any mail that is in the outgoing queue can be linked to a client.
Either it’s a hosted website that spams → we know who the site belongs to. The site may have been compromised and it would be useful to notify the owner.
Or it’s an outgoing transactional email (via authenticated SMTP) → we know who the client is, through authentication. It would be useful to inform the client that we have vaporized a few emails. They might be invoices, payment confirmations or password resets.
Or it’s a redirected mail. Depending on the email address that triggered the redirection, we also know which client will be affected by these missing emails. It might have been spam coming from outside. Or ham.
The problem is the forwarding, i.e., a received email that goes back through the outbound pipe to execute the forward. It’s in this pipe that the email gets deleted. OVH doesn’t want to be blacklisted because it emits spam.
If you use a real mailbox instead of a forward, all your emails will be there.
I’m dealing with a huge number of false positives, even within the same domain that has around forty MXPLAN and Email Pro addresses. When the anti‑spam filters the mail, the messages are never delivered—neither the sender nor the recipient gets notified—and adding trusted addresses to the whitelist via webmail doesn’t help; the filtering happens upstream. In short, this is highly penalising and wastes a lot of time and money for a business. MXPlan isn’t expensive, but it’s not worth much either.