Correct, but please see slight adjustments below - to try and make this tricky thing clear.
-
Thus, on a 14 dBm (ERP) limitation, you can’t have better than an antenna of 2.15 dBi if you’re transmitting power is already 14 dBm (ERP) (N.B. its often rounded down to 2dBi for easy calcs)
-
Thus, on a 14 dBm (ERP) limitation, you can’t have better than an antenna of 0 dBd if you’re transmitting power is already 14 dBm (ERP)
N.B. The ERP specification means that the power limit is expressed when using a dipole antenna.
N.B. A dipole antenna can be expressed as having 0 dBd gain or 2.15 dBi gain.
and as Jeff says, you can mitigate all this by adjusting for all the inevitable losses that occur (e.g. in the connectors and cable). Because of these losses, you can legally use an antenna of higher gain.
This is how you can calculate it.
e.g.
output power limit = + 14 dBm (erp)
loss in SMA connector = - 1 dB
loss in cable = - 2 dB (for 20 metres @ loss of 1db per 10 metres )
loss in antenna connector = - 1 dB
Antenna gain = + 4 dBd (antenna is advertised as 6dBi)
If you go through the calculation above, you see that the predicted power output from the antenna will still fit within the specified power limit because the losses match the antenna “gain”.
i.e. Even if an antenna with gain of 6dBi is used with RF power at the limit of 14dBm, the resultant field strength radiated by the antenna will still be within the limit.
Remember also, that the antenna is focussing the radiation and that the field strength is measured in the direction which gives the maximum strength.
For the calculation above, I used arbitrary dB figures for the connector and cable losses (e.g. 1dB per 10 metres). If you want more accurate “dB loss” values, you need to look up the spec sheets and check what the loss figures are for the frequency used e.g. 868Mhz
e.g. Have a look at this pdf link for a cable spec (check your downloads folder) which seems to suggest a dB loss value of about 1.5dB per 10m (i.e. 15dB per 100m)
https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=2ahUKEwi_8ZTP5P3cAhVOiRoKHfXKBJ0QFjAAegQIABAC&url=https%3A%2F%2Fwww.pasternack.com%2Fimages%2FProductPDF%2FRG400-U.pdf&usg=AOvVaw2F2LkOUIJW6NOrdeoaI20Y
Look for the “Electrical specifications by frequency”. What you see is that the cable loss value per metre is dependant on the frequency used - it increases with higher frequency.
Obviously for Europe and USA, we are looking for what the value would be at arround 900MHz).
If you have the correct specification sheet for the cable and the connectors you are using, and you know the antenna gain value (dBi or dBd), you should now be able to put it all together to calculate the dBm figures that would be output (by your antenna) at different lengths of cable.
Alternatively, you can use this on-line calculator
There are other considerations, but you can usually ignore them as long as you have a well matched antenna (50 ohm) and you are using a cable with the same “Characteristic Impedance” (50 ohm again),
The main thing to take away, is that antenna manufacturers always quote dBi figures - which is naughty really because power limits are almost always specified with ERP (and not EIRP).
What this means is that you have to “downgrade” the gain value to the correct figure by subtracting 2.15 dB from the dBi figure that the antenna manufacturer quotes.
Its no wonder that antenna manufacturers are quite happy for buyers to mistakingly believe that the more expensive the antennae, the more “magic smoke” the antenna has, and also the continued use of dBI to make their gain seem better (which just misleads the unwary 
)
The other takeaway is that distance gain is proportional to the square root of antenna gain
What this means is that it is realistically impossible to get an horizonally omnidirectional antenna with a reasonably focussed vertical radiation pattern (suitable for Lorawan) that achieves a distance gain greater than 4 times over a dipole positioned in the same place
Appologies for repeating the same thing