A simple, readable, and fast method would be to loop through the dates in date1, find the closest date in date2, and then replace that date in date2 with NaT so it won't be chosen again. That's what this demo does.
Inputs: two row vectors of dates in datetime format: dates1 & dates2
Outputs: nearestIdx(j) is the closest non-repeated date in date2 to date1(j); m is a nx2 matrix of paired dates.
See inline comments for details.
% Create two arrays of random dates (may contain repeated dates)
dates1 = datetime(2019,1,1)+days(sort(randi(364,1,100)));
dates2 = datetime(2018,12,28)+days(sort(randi(364,1,100)));
% Loop through the dates in date1
nearestIdx = nan(size(dates1)); % Pre allocation
dates2Temp = dates2; % Make a temp copy of dates2 for NaT replacement
for i = 1:numel(dates1)
[~, nearestIdx(i)] = min(abs(dates2Temp - dates1(i))); % find index of closest date
dates2Temp(nearestIdx(i)) = NaT; % replace that date with NaT
end
% Sanity check: all of the values in nearestIdx should be unique
assert(numel(unique(nearestIdx))==numel(dates1),'Sanity check failure: unique date matching error.')
% Match the dates (dates1 and dates2 must have same size)
m = [dates1.', dates2(nearestIdx).']
