The 500k for EQ2 was someones guess on a forum I googled, but I'm seeing a lot of people say 250 or 200k in response - all of that is taken prior to F2P. Then I applied this article to swell that by 300%.
So if it was 200k or 250k then it would be 600k or 750k after F2P, or if it were the 500k estimate then it would be 1.5M. They still have from the looks of it like 30-40 med-high pop servers in North America alone: that sounds impressive enough to support that (albeit fairly wide) range.
For Rift, they released that their revenue was over $100 million as of Jan 2012, at $15/mo for 12 months a year - that's 555,000 subscribers. They also said in November 2011 that they were still growing, so it's likely that was still true as of Jan 2012 - and was the reason they didn't want to experiment with F2P. Also looking at it now, that's before they entered Asian markets - which was likely at least net positive for them. However, http://mmodata.net/ suggests it dropped to 250,000 before it went F2P.
I can only seem to find articles that suggest a massive increase in players after moving to F2P - but the one I had found at work (at home now) had claimed a five-fold increase, which I had compared to the 200-250k estimates I was finding (ie. mmodata.net).
True, F2P players can't be counted as "subscriptions" - but Riot has suggested that they actually earn on average more than $15/mo/capita from being F2P. Now, MMO's might not be as successful as MOBA's for this model - but with so little data to work with, I think it's a safe bet (which is probably why so many companies are embracing it).
MMOdata.net is also my source for 1.5 million subscribers for SW:TOR before it went F2P. Combined with the 2 million increase in players mentioned here.