I input 34212 sepctra with several mgf files. Finally, it displays 33935 
predicitons, among the 33935 predictions, 32607 predictions have valid 
sequences, others (33935-32607=1328) are nan.

I understand that due to precursor/fragment tolerance and PTMs, there will be 
spectra without any valid prediction sequences,which shows nan (the 1328 
predictions).

However, why only 33935 predictions are displayed when 34212 spectra are input? 
What happened to the missing predictions?