PSD vs DFT: Why is this peak shifting?

The following image depicts output from a spectral analysis of (1) an auditory stimulus and (2) the recorded neural response to that stimulus using scalp electrodes, averaged over a group of subjects. The "Baseline" spectrum is the average of the neural response during the time period before the evoking stimulus was presented. The x axis is frequency in Hz and the y axis is the DFT (left) or PSD (right) value respectively.

What I'm wondering is: why does the location of the peak sometimes change in the response when comparing DFT to PSD? Is something going wrong with my calculation or plotting, or does this make sense that this could happen?

Here are the relevant chunks of code I am using:

if true

 %%SETTINGS
    fs = 20000;         % sampling rate (Hz)
    stimMs = 120;       % duration of stimulus (ms)
    prestimMs = 40;     % duration of prestimulus baseline recording period (ms)
    f = 350:750;        % vector of frequency values we are interested in analyzing (Hz)
    nWins = 5;          % number of desired analysis windows
    overlapRatio = 0.5  % ratio of overlap between windows

 %%DERIVE SOME MORE VARIABLES
    spms = fs/1000;                       % samples per ms
    wMs = floor(stimMs / (nWins + ...
    overlapRatio - nWins * overlapRatio));    % analysis window duration (ms)
    wSamps = wMs * smps;                  % # of samples per analysis window
    overlapMs = wMs * overlapRatio;       % duration of window overlap (ms)
    prestimSamps = prestimMs * spms;      % # of samples in prestim resp baseline
    w = blackman(wSamps);                 % define shape of window
    o = floor(overlapRatio * wSamps);     % # of samples of overlap between windows
    stimSamps = stimMs * spms;            % # of samples making up the stimulus waveform
    tStimOn = prestimSamps+1;             % response index when stim turns on
    tStimOff = tStimOn + stimSamps-1;     % response index when stim turns off

 %%GET THE DFT AND PSD VALUES 
    in1 = currResp(tStimOn:tStimOff);           % grab part of resp where stim is on
    in2 = currResp(tStimOn-wSamps:tStimOn-1);   % grab the prestim baseline
    [s1,~,~,p1] = spectrogram(in1,w,o,f,fs);    % DFT and PSD of resp 
    [s2,~,~,p2] = spectrogram(in2,w,0,f,fs);    % DFT and PSD of prestim baseline
    [s3,~,~,p3] = spectrogram(stim,w,o,f,fs);   % DFT and PSD of stim

 %%PLOTS
    for i=1:nWins

        % plot DFT
        y1 = abs(s1(:,i));     % response
        y2 = abs(s2(:));       % baseline
        y3 = abs(s3(:,i));     % stimulus
        subplot(nWins,2,2*i-1);
        plot(f,y1,f,y2,'-.',f,y3,'--');
        xlim([f(1) f(end)]);
        yMax=max(max(abs(s1)));
        ylim([0 yMax]);
        title([fName(1:end-4) ' Window ' num2str(i)],'Interpreter','none');

        % plot PSD
        y1 = p1(:,i);          % response
        y2 = p2(:);            % baseline
        y3 = p3(:,i);          % stimulus
        subplot(nWins,2,2*i);
        plot(f,y1,f,y2,'-.',f,y3,'--');
        xlim([f(1) f(end)]);
        yMax=max(max(abs(p1)));
        ylim([0 yMax]);
        title([fName(1:end-7) ' PSD Window ' num2str(i)],'Interpreter','none');

        % legend at the end
        if i==nWins
            lgd = legend('Response','Baseline','Stimulus','Location','none');
            set(lgd,'Position',[0.4,0.4175,0.2,0.1]);
        end

    end

end