The fundamental issue, whether you use PPMs or diaries, is sample size.  I believe that the GTA has 600 meters.  So it takes 6 meters to register a 1 share.  With so many stations licensed to the market, and a lot of them with shares 2%, the reality is that the predicted audience share for many stations is less than the standard deviation corresponding to the sample size. In other words, statistically insignificant. If your share really is in the double digits, like Radio 1,  CHFI or CHBM, then the ratings can be meaningful. 
I believe that Evanov has decried this problem for years, but given that Numeris is a not-for-profit funded by the member stations, the biggies don't want to increase the sample size because it potentially hurts them. 
The problem with diaries is that, based on my own experience, people don't fill them out promptly.  That means trying to recall the actual time duration hours (or days) later.  And of course, the potential for deliberate bias because there is no way of checking that the radio was actually turned on.  "If I was actually listening, I would have listened to CHUM".