Maybe the value of Z is wrong. It depends on the method you use to calculate Z and M.

May also be insufficient number of samples for analysis.

The answer is not hard. It is very simple:

It is simply that FishBase is a grossly unreliable source that should never be taken at face value. It is useful as an initial guide (and I use it frequently) but then you need to do some serious research, starting by checking the cited sources for the numbers that the website provides.

To go a bit deeper:

Quoting a single value for M in any given species is a dubious undertaking but a single value for F is simply absurd: F can be hugely different, between one population and another of the same species, or between one year and the next for the same population. Yet the compilers of FishBase have sought to fill in a lot of blanks, putting one number in each space for each species. Moreover, they have attempted to do so for every fish species in the biosphere -- an impossibly burdensome task, if anyone really cared about getting reliable values. In a forlorn attempt to achieve the unachievable, someone has grasped at whatever pieces of information they can find anywhere, no matter how absurd the outcome.

In the case you quote, I will guess that a value of M was drawn from one source, a value of Z from another and the computer simply found F by subtraction, without anyone checking for absurd results. If so, that should be clear from checking the cited sources. (One positive thing about FishBase: It does provide citations for its sources.)

In general: Use FishBase as a quick indication of some point or as an introduction to the literature, but do not rely on anything it presents.

usually, I use the following equation to estimate natural mortality proposed by Hoenig's (1983) ln(M) = 1.44 – 0.982*ln(maximum age)

You may give it a try.

One problem with John Hoenig's method is that, if you assume that natural mortality can be modelled with a single value of M, then you are assuming that mortality follows Baranov's exponential model. An inevitable consequence of that model is that the maximum observed age increases with sample size. John built his equation using numbers from studies that had each used sample sizes of a few hundred individuals. If you have the maximum age observed in a sample of about that size, John's method works as well as any. However, if your sample size is a few thousand (or a few tens of thousands), his equation yields wrong answers.

Worse, it is not the number of fish that were aged which matters but the size of a random sample which would have included the oldest fish in the actual sample. Hence, if someone picked out only the very largest fish observed and aged those, or if you troll the literature for the single largest reported age, your effective sample size can be orders of magnitude greater than the number of individuals actually aged.

Another problem with John's method is that it is not really an estimator of M at all but rather one of Z -- the observed maximum age depending on the mortality rate actually experienced by the fish, not one of two components into which we choose to partition the deaths. If your sample includes old fish from before a fishery developed, Z will approximate to M and the equation can work (though estimating effective sample size could get even more complicated). However, if you have samples from a long-established fishery, Hoenig's equation will give you some sort of estimate of Z. Indeed, his method sometimes "works" because the bias from using very large effective sample sizes causes an underestimate of Z that can end up being approximately equal to a plausible value for M. And that, while it can neatly give you a number to play with, can be dangerously misleading.

Under the right circumstances, Hoenig's method can be useful. So can Pauly's or (more simply) just an estimate of M from the reciprocal of von Beratalffy's k. Under the wrong circumstances, however, any method which purports to give you a value of M without the labour of a very extensive research program can lead you very, very seriously astray. In fisheries science (as in life in general) we can rarely get anything for nothing. Quick-and-dirty short-cuts can be useful sometimes, but you need to know what you are doing or you will fall into huge, embarrassing errors!

For more information on this topic, see my 2013 paper in Fish & Fisheries 15: 533-562.

I recommend you to compute the natural mortality rate using the updated version of Pauly (1980) formula, excluding the effect of temperature (T), which was fitted to a dataset of 218 species using non-linear least-square fitting described by Then et al. (2015) as follows:

M = 4.118K^0.73 ⋅ Loo^ - 0.331.