No, currently I can't help. Please post some relevant example data (by editing the original question, not as comment or answer), such that we can see, which of the inputs contains the NaNs and where.
Then an answer will be easy (or impossible): Most likely the NaNs can be simply removed before the interpolation.
[EDITED after your first comment]
Had = [1 3 4 6 nan nan 20 ];
time = [1 2 3 4 5 6 7];
good = ~isnan(Had);
had2 = [ 2.5 5 17]
time2= interp1(Had(good), time(good)', had2);
