I think one of the problems in comparing eras is that when one tries to account for some really important variable, like TOI, coaching and player use changes, influxes of new talent and expansion of teams, the method becomes somewhat subjective and then appears to "punish" one era over others (which is odd since something has changed why wouldn't something like scoring rates change?)
I think when comparing players offense, and how good they were at it, across eras one needs to account for and look at as many metrics as possible and somehow combine all the information into an objective look as is possible with subjective things.
I think too often people want to see all players on the same level across eras, more or less, so any metric that doesn't do this is seen as unreliable which may be faulty.
It might very well be that the differences in eras will produce different numbers in different seasons for the reasons of all the possible variable differences and just plain randomness at times.
The larger the sample the less chance there is for randomness and the closer any two players are in the era (or difference continuum) then the results are probably more "accurate".
Players at the extremes of the difference continuum, say McGee and Stamkos then it's extremely difficult to come up with just one single metric to measure how each player was compared to each other.
