All my level 20's are over 140 HP whether they have 18 CON or 45 CON or anything in between. I have 9 level 20+ characters, so I think I can judge how they're doing.
Summarizing:
1 Archer, 40 CON, 141 HP
1 Priest, 40 CON, 142 HP
1 Warrior, 38 CON, 158 HP
1 Warrior, 45 CON, 158 HP
1 Archer, 42 CON, 150 HP
1 Archer, 40 CON, 141 HP
1 Priest 49 CON, 131 HP
1 Archer, 45 CON, 175 HP
Those are my "high con" level 20 chars. my "low con" level 20 chars are:
1 Warrior, 24 CON, 146 HP
1 Warrior, 18 CON, 142 HP
Which isn't much less than the other warriors who had CON poured into them for the first ~10 levels.
You can judge and say whatever you desire. I wasnt trying to suggest otherwise. Same as I can disagree.
Your data.
1 Archer, 40 CON, 141 HP
1 Archer, 40 CON, 141 HP
1 Archer, 42 CON, 150 HP
1 Archer, 45 CON, 175 HP - More CON more HP
1 Warrior, 18 CON, 142 HP
1 Warrior, 24 CON, 146 HP
1 Warrior, 38 CON, 158 HP
1 Warrior, 45 CON, 158 HP - Same HP as 38 CON, otherwise more CON more HP
1 Priest, 40 CON, 142 HP
1 Priest, 49 CON, 131 HP - Less HP with more CON, anomaly.
Without knowing when and were the CON was spent, the data has little value though. The HP added is random, but theres quite a trend even in that small base, with only a couple of anomalies. More CON is more HP.
My characters at level 18, 30 CON all, CON was leveled almost consistently after a very small initial bit of random leveling (1-3 within each other), the low HP priest started with the lowest CON at 11, the rest 18:
Warrior 132
Warrior 128
Warrior 133
Archer 125
Archer 125
Archer 127
Priest 101
Priest 90
Looks very consistent to me. And although I said in this thread earlier that HP gain across classes was the same, I dont think so anymore.
Warrior 0.275
Archer 0.259
Priest 0.227
Thats the average as it stands for HP gained at level up for each CON point. Thats 20% more for warriors than priests.
The problem is imo that the random variation is high, from 0.11 HP/CON to 0.60 HP/CON, the levels to gain 20-25 low. Since it doesnt give alot of time to play out and show its face. Its easy to grab a low hp character and compare to a high hp character when the sample size is only small and the random factor of HP gain from low to high is huge.