You seem to be suggesting that 1 million ad impressions (or 20% of the site's total) happen when people are sending messages. Are you sure that's correct? Seems a little high for a classifieds site, where in my experience people tend to browse a lot.

That's really just an educated guess, based on the fact that I'm seeing more and more mobile traffic, and the average for mobile users is only 3-4 pages per session. So I'm thinking that many of them are just coming and finding what they want, without browsing around too much.

That could be a high estimate, of course... does the actual number change your your opinion on whether I should eliminate the page? If so, I might be able to find more data in Analytics, but it's hard to be too exact because there's a lot of chit-chat via PM... I can't just count the number of times the PMs were accessed.

I'm going to start from the end and work my way back...

Do you guys think that this change would increase the revenue

I doubt that whatever approach you choose you would be able to measure an impact and be attribute to the chosen approach. The question is valid, but I don't think you can evaluate the outcome based on earnings.

2.7 million of them were CPC bids, and 1.7 million were RPM bids

By RPM bids I assume you meant CPM bids. I am pointing this out because it is important to make the distinction between the two acronyms. CPM bids are cost per 1000 impressions and refer to the bid. RPM is revenue per 1000 impressions and has nothing to do with bid type. RPM is standardized measure of revenue such that bid type be CPC, CPM, AV-CPM etc... do not factor into it. This allows you to compare earnings across bid types, just what is called for here. To decide whether to favor clicks or impressions what you should do is compare the RPM of your CTR impressions (RPM shown on the clicks tab) against the RPM of the CPM impressions (RPM shown on the views tab). If RPM is higher for CPM then favoring more CPM impressions should generate more revenue. But again, it is not so simple because there may not be sufficient demand for one bid vs the other so adding more or less of one may not provide the optimal outcome. Theoretically if the AdSense market place is efficient (I assume it is pretty close, certainly not perfectly efficient) then the current mix should be the optimal one. (I digress....).

As mentioned above, I doubt you could measure the difference either way, but it is worth a check. Your niche may be such that one bid performs type much better than the other and so it is probably good to check it, but if the RPM figures are close in value I wouldn't worry about it.

A better user experience is key,..

The number one driver of revenue is traffic, more traffic == more options. Better ux typically means more traffic. I would approach this as UX questions. Based on what you describe here:

Currently, someone views a listing, clicks to send a message where they're taken to another page, then submits the message and are taken to a third page.

It seems like a painful user experience, and I think the AJAX approach is the best way to handle this. Specially in a mobile environment.

This would cut the pageviews from 3 to 1, and ad impressions from 9 to 3.

You can refresh the ads with AJAX provided that the page content changes substantively. You could take an approach, where an ad is shown at the first impression, then remains for various steps and then at the final step it refreshes when the final content is loaded on the page. Note, this would not work if your final step is just a screen that says thank you and bye. If you final screen is a fully refreshed and update view of the listing it would be acceptable.

Also, if an ad is visible while a user is watching a spinner while waiting for an upload to complete, the chance are pretty good that the user will focus on the ad more than the spinner. The AJAX approach has some advantages.