No, it is not.  But it is to good practice to prepare the sample with mobile phase, the baseline will be cleanner.

In such case, due to differences in polarity if layers are being separated, then what should we do?

In general, the sample must be soluble in the mobile phase.  If the sample is not soluble in the mobile phase, the polarity has to be adjusted to make them miscible.

Like, if the compound of interest is benzene....The mobile phase is Acetonitrile:Water. Can we use this mixture directly, to have a peak of Benzene?

You may make the benzene sample with acetonitrile.  If benzene concentration is low enough, it will soluble in Acetonitrile:Water.

You will see a peak of benzene in RPC mode.

Make the benzene sample with acetonitrile first, then dilute the sample with mobile phase (Acetonitrile:Water). 

Match the injection solvent with the mobile phase is always the best. If there ae problem of solubility, at least dissolve analytes in the same organic solvent of the eluent

It is 'really' a good idea to matrix match your standard and sample so they will give similar results.  Thus, the last dilution prior to injection is with mobile phase.  Remember, HPLC is a 'comparative' technology where you are comparing the results between the standard of known concentration with the sample!

Thanks everyone for the guidance

In RPC LCMS  i use the injection solvent (which should dissolve and misible with you analytes) close to the initial moblie phase. This is special true to those very polar and early eluting peaks. High percentage of organic solvent, say methanol or acetonitrile, will not favour the peak focus process in the LC column front and broad or spilt peak may happen to the early eluting peaks. Your analytes partition between the stationary and moblie phase. Hydrophliic polar compounds will spread and less condense to stationary phase if pure methanol injection solvent is used,end up in undesirable chromatographic peak.  Late eluting peaks is totally unaffected because they have high affinty to sationary phase.    

The short answer is no. You need to make sure that you get good refocusing on the head of the column (for example, polar compounds should be prepared in polar matrices) but if you have good refocusing then you can prepare the sample in a solvent other than the mobile phase.

My answer is NO. You have to prepare your extract in a non elutiong solvent mixture. For example, if you mobile phase is H2O/Méthanol 40/60, you can chose 50/50 or 60/40. With this recommendation, your compound refocusing on the head of your precolum.

While the answer is no, it does not make any sense to introduce an interference.  Usually, I make the final dilution in mobile phase so there are no sample matrix effects.