You also need to know how the average player does check a map. They somewhat got a boolean mind. Its good or not good.
The reasons why they found a map bad or good is because they experienced only a few parts. Those parts stay in their mind and other ones become forgotten. Also, the class they play decides alot and the players that are in it. If you have the 2fort IR type of players they generaly call it bad because most maps try to prevent the 2fort sniper spam. However, a team with mindless rushers and fun players that dont care about winning might call it good. They havent however won a stage. And again, the comments are useless because you need a serious team for it.
This makes finding feedback that is usefull hard and therefor is rare. The best players which provide usefull feedback are the ones who do play for fun but still try to win. Luckily the gamedays got many of those players which makes it a very usefull source for feedback.
Another thing is bugtesting. Best is to do this BEFORE an actual alpha/beta release and best is to find expert rocket/stickyjumpers for it. Any bug found during a test will couse players to change the playstyle even though they dont notice it (they will peek to that corner too often not looking at the place they normaly would). A single test often isnt enough also as people dont know the routes that well that it's somewhat automatic. Luckily these are minor balance issues and they will be found later when some servers use the map on their main rotation and not just for a test.
Hoodoo suffered from the last. It did have its playtests but with a limited type of players. It didnt have the randon 32 player IR type in its tests as there wasnt any server of those types willing to let their players download 70mb. As soon as it was released in a valve update the right feedback came to him and he already managed to massively improve stage 2 because of it. And im sure that before that map came in an update that he had alot of tests already (more than most maps).
The best indicator that your map is fun to play is when it starts to show up on servers more often. And best are the 32 IR servers for it. If your map works well on those servers they will work on allmost any other. (yes, respawn time tweaking is much easier to balance the map than editing the map). Hoodoo failed at that part and because of that recieved a very bad name.
I would like to point that sometimes you might want to restrict specifically what you want tested - for example if you're making a GravelPit style map, and the defensive team always defends A but the offensive team always goes B you're not really getting a good testing of how easy/hard it is for one team at B.
I disagree a bit with that. If the offenders allways go to B and the defenders to A you either have placed the spawns wrong where people dont notice A until B is capped (read cp_wolf) or it has been tested alot and the result is that A is harder to cap so red defends that one to make it even harder and guarantee a win and blue goes to B because thats some free extra time they recieve then. It would save them from some extra defence they would get if they capped A first.
in the first week of gravelpit defence was spreaded evenly also. Now allmost anyone goes B as defenders because its the easiest one to defend.
If you wait too long with fixing that you however are forced into locking 1 route or making them spawn facing that route where the other one gets somewhat forgotten so people will attack that point a bit more again. Then people will find out that the point indeed has changed and its easier again. The playtests in the next version are valuable again then.
Why do you think valve has over 100 iterations of a map? Because they want to fix those issues as fast as possible so people dont realy know which site is easier to attack and so spread out in a normal way.
Another thing, if people say the respawn times are too long then thats some valuable information. Dont ignore that as respawn times are something they will have to face all the time. There is however a chance of confusion. Sometimes the defenders have a respawn time of 30 seconds ingame but the map is balanced. In this case its often quite clear that they dont like the wait of 30 seconds even though its balanced. Sometimes its however 14 seconds and in that case there are 2 options they can mean. They dont like the waiting (in that case you can often ignore that because 14 seconds arent that long) or they are unable to defend the map. When they are unable to defend you however need to know how good the team actualy is in defending. Some others have already explained that part before and i dont bother to copy that.
The playtesting process is much more complex than most think.