Section 20.3.7 in 802.11n Standard states clearly about the subcarrier occupation in HT20/HT40/HT40+/HT40- channel mode:

… In the 20 MHz HT format, the signal is transmitted on subcarriers –28 to –1 and 1 to 28.

In the case of the 40 MHz HT format, a 40 MHz channel is used. The channel is divided into 128 subcarriers. The signal is transmitted on subcarriers –58 to –2 and 2 to 58. In the case of 40 MHz HT upper format or 40 MHz HT lower format, the upper or lower 20 MHz is divided into 64 subcarriers. The signal is transmitted on subcarriers –60 to –4 in the case of a 40 MHz HT lower format transmission and on subcarriers 4 to 60 in the case of a 40 MHz HT upper format transmission.

The table below summarizes the quoted text.

Channel Mode Bandwidth(Eff.) Number of Subcarriers Subcarrier Index
Legacy 20Mhz 20(16.25)MHz 52 -26~-1 1~26
HT20 20(17.5)MHz 56 -28~-1 1~28
HT40- 20(17.5)Mhz 56 -60~-4
HT40+ 20(17.5)Mhz 56 4~60
HT40 40(35.625)MHz 114 -58~-2 2~58

See! HT20/HT40-/HT40+ all occupy 56 subcarriers, but their subcarrier indeces are different!

Does the Subcarrier Index Matter ?

YES, especially for phase measurement. Both baseband SFO (Sampling Frequency Offset) and PDD (Packet Detection Delay) contribute a linear slope error term to the phase. However, in HT20 and HT40 mode, the 0-th subcarrier, not existent but can be interpolated, is NOT affected by SFO and PDD. Moreover, the linear regression based SFO and PDD elimination technique is relatively accurate.

However, in HT40+/- mode, ALL subcarrier experience much stronger distortion. The 0-th subcarrier, far from working subcarriers, cannot be interpolated. Further, the linear regression is less accurate without the 0-th subcarrier being reference.

Conclusion

Use HT20 or HT40 mode! For HT40+/- mode, no scenario is find which prefers the more biased phased.