RBNGate spots the callsign actually heard by the skimmers (with or without the "/P" as the actual case may be). Although the skimmers do occasionally bust a callsign, it is highly unlikely that a skimmer would either append a "/P" when none was sent, or omit the "/P" entirely if it was sent.
Since October 2013, I've been maintaining a log of every spot ever posted to SOTAWatch, and that log can be imported into my local database for analysis. Here is some interesting data:
Since October 20, 2013 W4DOW has posted a total of 1,039 spots. Of those, 1,016 (97.8%) of his spots have included the "/P". Of the 23 spots that Dow posted without the "/P", 10 of them already had portable suffixes such as "/W1" which the activator likely sent, and for which appending a "/P" would be superflous. In one other instance he intended to include "/P" but had a typo in his spot ("VA2VL.P"). If we exclude his 10 spots with portable suffixes and add his typo spot to his /P total, it means that 1017/1029 = 98.8% of his spots are of the "/P" variety.
During that same period, RBNGate posted 26,850 spots and only 17,504 of them (65.2%) included a "/P" designator. RBNGate only posts CW and digital mode spots, but if we assume that activators using one of the phone modes have an equivalent preference for using or not using "/P", and that human spotters are reasonably accurate, we can see a stark contrast. In summary:
W4DOW - 98.8% of spots contain "/P"
Everyone else - 65.5% of spots contain "/P"
Free SOTA Spot Monitor Software + RBNGate FAQ: