I'd agree with that as a casual definition, but I think what the algorithm considers 'competitive' is more like 1 & 2.
Just to be clear, there is no specific 'competitiveness' calculation in the algorithm per se, but what the algorithm does is correlate between score and expected difference in DNTRP. My estimate (who the heck knows for sure) is that a score of 1 & 2 (maybe even 1 & 1) translates to an expected DNTRP difference of 0.5 (i.e. one full level). Just based on experience (observed records of people being bumped and so on), and the various statements USTA drops on us from time to time (e.g. what ranges of scores to reasonably expect within level).
In the case you mentioned, if the player in question was getting 2-3 games per match against *top* 3.5's, that would explain the bump up.